# Emoti: the recruiting agent that shows its work > Emoti is a recruiting agent. It finds candidates, writes personal outreach, runs first-round voice screens and takes interview notes, and it shows the evidence behind every ranking so recruiters make the call. Canonical site: https://emoti.ai. Published by Emoti Research. ## What Emoti does ### Sourcing: Sourcing from a sentence Write who you need the way you would tell a colleague. Emoti reads it into criteria, checks every profile against each one, and shows its evidence. - **You describe the role.** Title, level, place, the experience that matters and the skills you would accept instead. Paste a job description if you have one. - **Emoti reads it into criteria.** Each requirement becomes a criterion you can see and edit, marked as a must-have or a nice-to-have. - **Every profile is checked against every criterion.** Each check comes back met, partly met, missing or unknown, with the line of the profile it is based on. - **You get a ranked shortlist with its reasons.** Must-haves count double, and missing one caps the score at 54 out of 100, so near-misses stay out of the strong-fit range. Shortlist, pass, or find people like the best match. ### Outreach: Outreach you sign off on Sequences that open with a line from each candidate's own work, drafted by Emoti and sent only after you approve them. - **You choose the people and the sequence.** Add shortlisted candidates to a sequence of emails and LinkedIn messages with the gaps you want between steps. - **Emoti drafts a first line for each person.** Each line comes from something in their own profile: a project, a talk, a change of role. It never guesses at things they did not share. - **You approve, edit or rewrite every line.** Nothing is sent until a recruiter approves it, and you can change any draft before you do. - **Emoti sends inside your limits.** Messages go out on weekdays inside your sending window and daily limit, and any reply stops the sequence for that person, whether it says yes, not now or no. ### Voice screens: First screens, any hour Candidates take a short voice interview when it suits them. The same questions for everyone, scored against your rubric, reviewed by a person. - **You set the questions and the rubric.** Start from the role's brief, edit the questions, and say what a strong and a weak answer sound like. - **Candidates take it when it suits them.** The invite and the screen both say an AI interviewer asks the questions and a person reviews the answers. Candidates confirm that before starting, and can ask for an adjustment or a person instead. - **Each answer is rated with its evidence.** Every rating sits next to a summary of the answer it rates, so a reviewer can disagree with any of them in seconds. - **A person decides who moves on.** Emoti sorts the results; it does not reject anyone. A recruiter listens and makes the call. ### Interview notes: Notes that point to the moment Emoti writes interview notes in your template and links every point to where it was said, so the debrief starts from evidence, not memory. - **Emoti joins the call, with consent.** It joins the interviews you choose and tells everyone on the call that notes are being taken. - **Notes arrive in your template.** Screen, technical, portfolio, behavioural or debrief: the notes follow the sections your team already uses. - **Every point links to its moment.** Open any point in the notes to read exactly what was said, and when. - **Ask the interview anything.** Ask what the candidate said about notice, pay or a project, and get their words back with timestamps. ## Pricing Every new workspace starts with a 14-day Team trial; no card is needed. Source: https://emoti.ai/pricing | Plan | Price | Contact reveals per seat a month | Voice-screen minutes a month | For | | --- | --- | --- | --- | --- | | Free | Free | 25 | 60 | For one recruiter trying Emoti on a live role. | | Team | $79 a seat a month ($63 billed yearly) | 400 | 1,000 | For a recruiting team running several roles at once. | | Business | $139 a seat a month ($111 billed yearly) | 1,000 | 5,000 | For hiring at volume with admin controls and an audit trail. | | Enterprise | Custom | Unlimited | Unlimited | For large teams that need a security review, custom terms and invoicing. | ## Questions ### What is Emoti? Emoti is a recruiting agent. It finds candidates from a plain-language brief, writes outreach for you to approve, runs first-round voice screens and takes interview notes, and it shows the evidence behind every ranking. ### Does Emoti decide who gets hired? No. Emoti ranks, drafts and summarises, and people make every decision. It never rejects a candidate on its own, and every screen result is reviewed by a person before anyone moves forward or is turned down. ### How does Emoti rank candidates? It reads your brief into criteria and checks each profile against them as met, partly met, missing or unknown. Must-haves count double, a missing must-have caps the score at 54 out of 100 and a partly met one at 74. The research paper on evidence-first matching explains the method and its limits in full. ### Does anything go out without my approval? No. Emoti drafts the first line of each message from the candidate's own work, and nothing sends until a recruiter approves it. Sending also respects a daily limit and a sending window you set, and any reply stops the sequence for that person. ### Do candidates know when they are talking to AI? Yes. The invite and the welcome page both say an AI interviewer asks the questions and that a person reviews the answers, and candidates confirm it before starting. They can ask for an adjustment or a conversation with a person instead. ### Which AI models does Emoti use? Emoti sends model requests through its own gateway. An admin assigns a model, with fallbacks, to each kind of task, and can see which provider and model handled each request. If no model is set up, built-in rules answer and Emoti labels the answer that way. ### What does Emoti cost? There is a free plan for one recruiter. Team is $79 a seat a month and Business is $139, with about 20% off when billed yearly. Every new workspace starts with a 14-day Team trial, and no card is needed. ### Can I try it before talking to anyone? Yes. Open the demo workspace to try every feature on sample roles, candidates and interviews, or start a free workspace of your own. ## Research paper: Designing first-round voice screens candidates can trust > Emoti's first-round voice screens tell candidates that an AI interviewer asks the questions, ask everyone the same questions, rate each answer against the team's rubric, and leave every advance or reject decision to a recruiter. The design still has gaps, including rubrics anchored only at the ends, scorecards that show summaries rather than the answers themselves, and no setting for how long recordings are kept. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Research paper, version 1.0. For talent leaders. Canonical: https://emoti.ai/research/trustworthy-voice-screens ### Abstract An AI voice screen asks every candidate for a role the same short set of questions, in their chosen language, at a time that suits them, and rates each answer against a rubric the hiring team wrote. Emoti's design rests on five commitments: tell candidates plainly that an AI interviewer asks the questions and that people review the answers; ask everyone the same questions; rate answers against written standards and show the rating with a summary; offer ways to adjust the format; and leave every advance or reject decision to a recruiter. This paper describes how each commitment appears in the product, where the design falls short of it today, and the risks that a voice format adds, including speech recognition errors that differ between groups of speakers, speech differences and interview anxiety. It reports no accuracy or fairness results, because we have not measured them yet. ### What is a first-round voice screen in Emoti? It is a short structured interview run by an AI interviewer: a fixed list of questions, each with a time guide and a rubric, which a candidate takes by voice through a link at a time that suits them. A recruiter builds the screen from these parts: | Part | What the recruiter sets | | --- | --- | | Questions | The wording, in the order they are asked | | Reveals | What a good answer tells the team | | Strong and weak | What a strong answer includes and what a weak one looks like | | Time | A guide of 1 to 20 minutes per question | | Follow-ups | On or off for each question | | Camera | Required, optional or off | | Languages and voices | Up to ten languages, each with a choice of four voice styles | | Welcome page | A headline and a message in the team's own words | A new screen starts with two general questions to edit: one about a piece of work the candidate is proudest of, and one about what they want next and when they could start. The builder shows the total length and comments on it: under 10 minutes is short, up to 25 minutes is comfortable to finish in one sitting, and anything longer is "a lot to ask before a first conversation". A screen can't be published until every question has wording. Candidates are invited by a shared link or by email, with one reminder. The format borrows from research on structured interviews, which finds that asking everyone the same job-related questions and rating answers against set standards predicts job performance better than unstructured conversation. A 2022 re-analysis ranked structured interviews first among common selection methods, with a mean validity of .42, and also found wide variation between studies (Sackett et al., 2022). That research is the reason for the format. It is not evidence that any particular AI screen works, and we have not validated ours. ### How are candidates told what is happening? Before the first question, the welcome page tells the candidate how long the screen takes, what the camera rule is, which language and voice they will hear, and who listens: an AI interviewer asks the questions, and people at the company review every answer before any decision. The page also says the candidate can pause between questions, and its footer says the interview is run by Emoti. The default welcome message, which teams can rewrite, says the same thing in a warmer voice: > Thanks for applying. An AI interviewer will ask a few questions about your work. Answer as you would with a person, and take your time. Someone on our team reviews every answer before any decision is made. The invitation email says an AI interviewer asks the questions, that a person reviews every answer, and that candidates can ask for an adjustment before they start. On the welcome page, "Start the screen" stays unavailable until the candidate ticks a box confirming that an AI interviewer asks the questions, that their answers are recorded and transcribed, and that people review them before any decision. Some laws ask for more. Illinois's Artificial Intelligence Video Interview Act requires employers that use AI to analyze recorded video interviews for Illinois-based positions to tell applicants beforehand, explain how the AI works and what general types of characteristics it uses, and obtain consent before the interview. New York City's Local Law 144 requires employers using an automated employment decision tool to assess candidates who live in the city to give notice 10 business days before use, including instructions for requesting a reasonable accommodation. Those duties sit with the employer, and this paper is not legal advice. ### Does everyone get the same questions? Within one version of a screen, yes: the same questions, in the same order, with the same time guides. The differences are the language a candidate chooses and, where they are turned on, follow-up questions. Follow-ups are a real tension. They can help a nervous or brief candidate give a fuller answer, and they also make interviews less alike. A major review of the research recommends using only planned, neutral probes that don't cue the desired answer, probing equally across applicants, and probing only when an answer is clearly deficient or off track, or the applicant is shy (Levashina et al., 2014). Emoti's follow-ups are switched on per question and used "when an answer is thin". In version 1.0 they are not written in advance, which falls short of that recommendation. Screens can also change while they are live. Edits apply to anyone who starts the screen afterwards, so candidates who started earlier answered the earlier version. When a question is removed, the scorecards of people who answered it say "This question has since been removed from the screen." Teams comparing a group of candidates should avoid editing a screen partway through the group. ### How are answers scored, and who decides? Each answer is rated on a four-point scale (weak, mixed, good, strong) against that question's rubric, with a one-line summary, and the screen's overall score is the average rating rescaled from 0 to 100. A recruiter reads the scorecard and makes every decision. The rescaling is simple: all strong answers give 100, all weak answers give 0, and five answers rated strong, strong, good, strong and good give 87. The overall score uses the same bands and words as fit scores elsewhere in Emoti: 80 and above strong, 60 to 79 good, 40 to 59 partial, below 40 weak. The scorecard labels the ratings "Rated by Emoti against your rubric" and marks the summary as written by Emoti. What is rated is the content of the answer, against the rubric. The camera never affects a score, and nothing in the design infers emotion or personality from a face or a voice. That matters beyond good practice: the EU AI Act prohibits AI systems that infer emotions in the workplace (Article 5(1)(f)), and the European Commission's guidelines on prohibited practices treat hiring as part of the workplace for that rule. #### Two weaknesses in the rubric design - **Anchors only at the ends.** The builder asks what a strong answer and a weak answer look like, but not the two points between them. One study summarized by Levashina et al. (2014) found that ratings on a five-point scale with an anchor at every point were more resistant to disability bias than ratings on a scale anchored only at the highest and lowest points. - **The summary sits where the answer should.** In version 1.0 the scorecard shows Emoti's rating and a one-line summary of each answer, but not the candidate's own words or recording beside it. A reviewer who wants to disagree with a rating needs the answer itself. Until it is shown, "people review every answer" depends on reviewers going beyond the summary. #### The recruiter's decision A completed screen waits as "To review" until a person chooses Advance or Reject, and either choice can be moved back to review. There is no pass mark and no automatic rejection. The candidate list shows people waiting for review first, highest score first, then people who have started or been invited, then decisions already made. Sorting by score helps a recruiter start somewhere, and it also invites anchoring on the number, which is one more reason the answer belongs on the scorecard. Keeping a person on the decision also reflects the law in some places. In the EU, the GDPR gives people the right not to be subject to a decision based solely on automated processing that significantly affects them, with limited exceptions (Article 22). ### How does the design handle accents, speech differences and nerves? Partly. Candidates can choose a language, hear a question again, pause between questions and take the screen when they choose, and the rubric rates what was said rather than how it sounded. The design does not yet correct for speech recognition errors, offer a written or human alternative inside the flow, or let a candidate ask for more time. The risks are real: - **Accents and dialects.** Rating spoken answers depends on a transcript, and speech recognition makes more errors for some speakers than others. In a study of five commercial systems, the average word error rate was 0.35 for Black speakers and 0.19 for white speakers (Koenecke et al., 2020). - **Speech differences.** People who stammer, have a voice disorder or speak through assistive technology may be misheard more often or need more time. We have not measured how Emoti's transcripts or ratings behave for them. - **Nerves.** Across studies, candidates who report more interview anxiety tend to receive somewhat lower interview ratings, with an overall correlation of −.19 in one meta-analysis (Powell et al., 2018). Speaking to a machine may calm some candidates and unsettle others. | Need | What version 1.0 offers | What it doesn't offer yet | | --- | --- | --- | | Language | Up to ten languages, chosen by the candidate | Checks that ratings agree across languages | | Hearing a question | "Hear it again" before answering | Captions or a written copy by default | | Time | Visible time guides and pauses between questions | A setting for extra time | | Format | Audio-only and camera-optional screens | A written alternative in the flow | | A person instead | Whatever the employer arranges outside Emoti | A built-in request for a human interviewer | Employers in many places must make adjustments. In the US, the Americans with Disabilities Act counts not making reasonable accommodations to the known limitations of a qualified applicant with a disability as discrimination, unless the accommodation would impose an undue hardship. In Great Britain, the Equality Act 2010 places a duty on employers to make reasonable adjustments. The welcome page carries a link for this: a candidate can ask for more time, written answers instead of speaking, a conversation with a person, or something else, with a note for the team. We still recommend that every team says in its own welcome message how it handles those requests. ### Limitations - We have not validated Emoti's ratings against trained human raters or against later job performance. - Ratings depend on transcripts, and transcripts are less accurate for some speakers. - Rubrics anchor only the strong and weak ends of the scale. - The scorecard shows summaries, not the answers themselves. - Follow-ups are not scripted in advance, so they can differ between candidates. - A live screen can be edited partway through a group of candidates. - Adjustment requests are not yet collected into their own queue for recruiters. - There is no setting for how long screen recordings are kept. ### What we have not measured yet We have not yet measured: - Agreement between Emoti's ratings and trained raters using the same rubric. - Whether ratings differ by accent, dialect, language or speech difference when the content of an answer is the same. - Whether advance rates differ by group, using a measure such as the four-fifths guideline. That needs outcome data from employers. - How candidates experience the screen, and whether completion rates differ by group. - Whether follow-up questions make ratings more or less consistent. - Whether screen scores predict later interview results or job performance. ### Key terms - **AI voice screen**: A first-round interview in which an AI interviewer asks a fixed set of questions by voice and each answer is rated against a rubric the hiring team wrote. In Emoti, a recruiter reviews the results and makes every decision. - **Rubric**: A written guide for rating answers: for each question or competency, what a strong, middling and weak answer contains. - **Scorecard**: The record of one interviewer's or one screen's ratings for a candidate, competency by competency or question by question, with the evidence for each rating. - **Structured interview**: An interview in which every candidate is asked the same job-related questions and every answer is rated against the same written standard. - **Reasonable adjustment**: A change to a hiring process or job that removes a disadvantage for a disabled person, such as extra time, a different format or a human interviewer. US law calls it a reasonable accommodation. ### Questions #### Who writes the questions and the rubric in an Emoti voice screen? The hiring team. Emoti starts each new screen with two general questions to edit, and the team writes what each question reveals and what strong and weak answers look like. #### How long should a first-round voice screen be? Emoti's builder treats up to 25 minutes as comfortable to finish in one sitting and flags anything longer. Shorter screens leave less room for follow-up questions. #### Does an Emoti voice screen use video? Only if the team chooses. The camera can be required, optional or off, and scores never depend on it. #### Can a team set a pass mark so low scores are rejected automatically? No. Version 1.0 has no pass mark and no automatic rejection. Every advance or reject is a person's choice. ### Sources - Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range. Journal of Applied Psychology, 107(11), 2040–2068. https://doi.org/10.1037/apl0000994 - Levashina, J., Hartwell, C. J., Morgeson, F. P., & Campion, M. A. (2014). The structured employment interview: Narrative and quantitative review of the research literature. Personnel Psychology, 67(1), 241–293. https://doi.org/10.1111/peps.12052 - Koenecke, A., Nam, A., Lake, E., Nudell, J., Quartey, M., Mengesha, Z., Toups, C., Rickford, J. R., Jurafsky, D., & Goel, S. (2020). Racial disparities in automated speech recognition. Proceedings of the National Academy of Sciences, 117(14), 7684–7689. https://doi.org/10.1073/pnas.1915768117 - Powell, D. M., Stanley, D. J., & Brown, K. N. (2018). Meta-analysis of the relation between interview anxiety and interview performance. Canadian Journal of Behavioural Science, 50(4), 195–207. https://doi.org/10.1037/cbs0000108 - Illinois General Assembly (2019). Artificial Intelligence Video Interview Act (820 ILCS 42). Illinois Compiled Statutes. https://www.ilga.gov/legislation/ilcs/ilcs3.asp?ActID=4015&ChapterID=68 - NYC Department of Consumer and Worker Protection (2023). Automated Employment Decision Tools: Frequently Asked Questions. City of New York. https://www.nyc.gov/assets/dca/downloads/pdf/about/DCWP-AEDT-FAQ.pdf - European Parliament and Council of the European Union (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2024/1689/oj - European Commission (2025). Commission publishes the Guidelines on prohibited artificial intelligence (AI) practices, as defined by the AI Act. European Commission. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act - Dushi, D. (2026). Red Lines under EU AI Act: Unpacking the prohibition of emotion recognition in the workplace and education institutions. Future of Privacy Forum. https://fpf.org/?p=248772 - European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Article 22. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj - United States Congress (1990). Americans with Disabilities Act, 42 U.S.C. § 12112. United States Code, via govinfo.gov. https://www.govinfo.gov/content/pkg/USCODE-2010-title42/pdf/USCODE-2010-title42-chap126-subchapI-sec12112.pdf - UK Parliament (2010). Equality Act 2010, section 20: Duty to make adjustments. legislation.gov.uk. https://www.legislation.gov.uk/ukpga/2010/15/section/20 ## Research paper: Evidence-first candidate matching > Emoti ranks candidates by checking each criterion in a recruiter's brief against what a profile actually says, and it never shows a score without that evidence. Each criterion is marked met, partly met, missing or not enough information; must-haves count double, and a missing must-have caps the score at 54 out of 100. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Research paper, version 1.0. For recruiters. Canonical: https://emoti.ai/research/evidence-first-matching ### Abstract Emoti ranks candidates against a recruiter's brief by turning the brief into a short list of named criteria, checking each criterion against what a profile actually says, and combining the results with a fixed formula. Every score is shown with its evidence: one mark per criterion, labeled met, partly met, missing or not enough information, with the line of profile text that decided it. This paper describes how briefs become criteria, how each criterion is checked, how must-haves cap a score, how the one-sentence explanation is assembled, and how look-alike searches build criteria from a single profile. It ends with the method's known weaknesses, including its reliance on keywords, its treatment of information a profile leaves out, and the bias that profiles themselves carry, and with the measurements we have not yet made. ### Why does Emoti show evidence before a score? A score on its own can't be checked, and a ranking that can't be checked gets either ignored or trusted too much. Emoti's rule is that no candidate's score appears without the evidence behind each part of it. Ranking decides who a recruiter reads first and who they never reach. That is a real decision, even when nobody is rejected. So the matcher is built to be read: every criterion is named, every result has a state and a line of evidence from the profile, and the score is a fixed formula over those states. The same profile and the same criteria always produce the same score. A language model can help read the recruiter's brief into criteria. It never scores people. Scoring is done by the same deterministic rules in every case, which is what makes each number explainable. ### How does a brief become criteria? Emoti reads the brief into a short, ordered list of criteria, each marked as a must-have or a nice-to-have, and shows them as editable chips above the results. The order is role, location, experience, skills, industry, then other signals. Two readers can produce the list: - **The model reader.** When a model is configured, Emoti's model gateway runs the request on whichever model the workspace admin assigned to parsing, with fallbacks in order, and the model is asked for three to eight criteria in that order, using only the matcher's own vocabulary for roles, places, skills and industries. Emoti then drops anything the matcher can't check, and any duplicates. If nothing usable survives, or the call fails or is declined, the rules reader runs instead. - **The rules reader.** Without a model, Emoti matches the brief against fixed word lists: eight role families, seniority words such as senior, staff and lead, a list of cities, countries and regions, the word remote, a pattern for years of experience, a dictionary of skills and their aliases (k8s becomes Kubernetes), a list of industries, and phrases such as "led a team". The page always says which reader produced the criteria: "Criteria read by the AI model" or "Criteria read by Emoti's built-in rules." | Kind | Example label | Default | What the matcher compares | | --- | --- | --- | --- | | Role | Senior backend engineer | Must-have | Role family and seniority on the profile | | Location | Based in Berlin or Europe, or remote | Must-have, unless remote is allowed | City first, then region | | Experience | 6+ years of experience | Must-have | Total years on the profile | | Skill | Kafka | Nice-to-have | Listed skills, then related skills | | Industry | Experience in payments | Nice-to-have | Industries of past employers | | Signal | Has led a team | Nice-to-have | Signals on the profile | #### Defaults and editing The rules reader, and the instructions given to the model, follow the same defaults. The role, a stated location and a minimum number of years start as must-haves. Skills, industry and signals start as nice-to-haves. If the brief allows remote work, location becomes a nice-to-have. A recruiter can switch any criterion between must-have and nice-to-have, remove it, add a new one in plain words, or edit the brief and rebuild the list. Shortlist decisions survive a rebuild. #### Look-alike searches A look-alike search builds its criteria from one person's profile instead of a brief, then ranks the pool with the same scorer. "People like this one" becomes a set of named, editable criteria rather than a hidden similarity score. From the chosen profile Emoti takes: 1. The same role family, as a must-have. 2. The same region, as a nice-to-have. 3. At least their years of experience minus two, with a floor of two years, as a must-have. 4. The first four skills listed on their profile, each as a nice-to-have. 5. The first industry on their profile, as a nice-to-have. Nothing else is copied: not the person's name, school or employer. The criteria appear as chips like any other search, so a recruiter can remove a skill that was incidental or add one that matters. ### What are the four evidence states? Each criterion gets one of four states, and each state has a fixed value: met is worth 1, partly met 0.5, not enough information 0.25 and missing 0. | State | Value | Example evidence | When it is used | | --- | --- | --- | --- | | Met | 1 | Used at Example Co. | The profile shows the requirement | | Partly met | 0.5 | Has TensorFlow instead | A close substitute or a near miss | | Not enough information | 0.25 | Not mentioned in profile | Only for signal criteria in version 1.0 | | Missing | 0 | No PyTorch in profile | The profile doesn't show it | Partly met has specific triggers. A role in an adjacent discipline counts as partial, for example data engineering when the brief asks for machine learning. So does a candidate at mid level when the brief asks for senior, a person in another city in the same region (with "open to relocating" added when their summary says so), someone one year short of the minimum, and a related skill, such as TensorFlow for PyTorch or Rust for Go. The evidence line is written from the profile itself: the title and employer, the city, the number of years, the employer where a skill was used, or the substitute skill found. It is short on purpose. It says what the matcher saw, not what it concluded about the person. ### How is the score calculated? Must-haves count double, the values are added up and divided by the most a person could score, and then two caps apply. Each criterion has a weight of 2 if it is a must-have and 1 if not. The score is 100 times the sum of each state's value multiplied by its weight, divided by the sum of the weights, rounded to a whole number. Then: - If any must-have is missing, the score is capped at 54. - Otherwise, if any must-have is only partly met, the score is capped at 74. Scores are shown with a word: 80 and above is a strong fit, 60 to 79 a good fit, 40 to 59 a partial fit and below 40 a weak fit. A missing must-have therefore keeps a person in the partial band at best, and a partly met must-have keeps them out of the strong band. #### A worked example The caps exist because a weighted average alone lets several nice-to-haves outweigh a missing must-have. Take a search with three must-haves (role, location, years) and three nice-to-haves (two skills and an industry), so the most a person can score is a weight of 9. | Candidate | What the strip shows | Before the cap | Score | | --- | --- | --- | --- | | A | Every must-have met, one nice-to-have missing | 89 | 89, strong fit | | B | Location partly met, everything else met | 89 | 74, good fit | | C | Years missing, everything else met | 78 | 54, partial fit | | D | Every must-have met, every nice-to-have missing | 67 | 67, good fit | Candidate B ranks above candidate D, even though D meets every must-have and B does not. The caps keep near misses out of the higher bands; they do not guarantee that everyone who meets every must-have ranks above every near miss. To make that visible, the results page states how many people meet every must-have, and the strip shows which must-have each person misses. Ties are broken by an internal identifier, an order that is stable but arbitrary. #### The one-sentence explanation Each result also carries one or two plain sentences built from the same results as the score: what the person does now, in their profile's own words, then how many criteria they meet, then up to two gaps. When nothing is missing, it names up to two partly met criteria instead. > At Example Co.: owns the pricing models used in every quote. Meets 5 of 6 criteria. Missing Experience in logistics. The sentence is assembled from the results, not written by a model, so it can't add a fact the strip doesn't show. Its weakness is the same as its strength: it is only as good as the profile text it quotes. ### Where does the evidence appear? In three places: the fit strip on every result, the "How they match" list in each profile, and a coverage panel that shows how many people in the pool meet each criterion. The fit strip has one segment per criterion, in criterion order. A met criterion is a solid segment, partly met is half filled, missing is hollow and not enough information is pale. The fill level carries the meaning, so the strip reads without relying on color. Hovering a segment shows the criterion, its state and its evidence line. For screen readers the strip is announced as a tally, such as "Meets 4 of 6 criteria". The profile repeats the same results as a list, with an icon for each state and the evidence under each criterion. The coverage panel shows, for each criterion, the share of the pool that meets it. When fewer than 8 percent of people meet a must-have, it says so and suggests making that criterion a nice-to-have. Results arrive in batches of ten. Shortlisting or passing on a person moves them out of the list to review. In version 1.0 these decisions do not change the criteria or the weights. The ranking changes only when the criteria change, which keeps every change visible. ### Limitations The method is simple on purpose, and its weaknesses come from the same choices. - **Evidence is keyword evidence.** A skill counts as met when it appears in the profile's skill list, and as "used at" an employer when the first word of the skill appears in that role's description. Listing a skill is not proof of depth, and a skill described in plain words without its name is missed. - **The vocabulary is closed.** There are eight role families, a fixed list of places, and fixed lists of skills and industries. A brief for a role outside those families, such as a recruiter, gets no role criterion at all. The model reader is held to the same vocabulary so that both readers can be scored the same way. - **The rules reader misreads ordinary words.** In a test brief saying a backend engineer would "lead the payments pipeline and help us go live in March, London", the rules reader produced a "Lead backend engineer" title, a Go skill from "go live", and North America as an allowed region from the word "us". The chips show these mistakes, but only to someone who reads them. - **Either-or requirements can't be expressed.** "Go or Rust" becomes two separate skill criteria. Making both must-haves would demand both. - **Unknown is not missing, but version 1.0 often treats it as missing.** Only signal criteria can be marked "not enough information". A skill, industry or place that the profile doesn't mention is marked missing, even though the person may simply not have listed it. The evidence line says "No PyTorch in profile", which is accurate, but the score treats silence as a gap. - **Profiles undersell some people.** People who write short profiles, describe their work without product names, change careers or have gaps look weaker to any matcher that reads profiles. Years of experience is a rough proxy for ability and can count against capable people with shorter or interrupted careers. - **Profiles carry bias in what they contain.** Titles, employers, conference talks and open-source work reflect access and visibility as well as ability. Signals such as "conference speaker" favor people with the time and networks to speak in public. The matcher doesn't read names, photos, age or gender, and it builds no criteria from schools, but that doesn't remove bias carried in the fields it does read. Look-alike search can reproduce the profile of an existing team, including its blind spots. - **The caps don't guarantee order.** As the worked example shows, a person who meets every must-have but few nice-to-haves can rank below a near miss. With three must-haves and six nice-to-haves, meeting only the must-haves scores 50, below the 54 cap of someone who misses one must-have and meets everything else. - **Partial rules are coarse.** One year short is partial and two years short is missing. Adjacent disciplines come from a fixed table. ### What we have not measured yet We have not measured whether the method finds the right people. Specifically, we have not yet measured: - How closely the ranking agrees with experienced recruiters who judge the same profiles independently. - Whether people ranked highly go on to pass screens, receive offers or do well in the job. - Whether shortlist rates differ by sex, race or ethnicity. Checking this needs demographic data that Emoti doesn't collect, compared with a method such as the four-fifths guideline in the US Uniform Guidelines on Employee Selection Procedures (29 CFR 1607.4(D)). - How often the model reader and the rules reader produce different criteria from the same brief, and which version recruiters keep. - How often "used at" evidence reflects real use of a skill rather than a word match. - Whether recruiters would decide differently if people who meet every must-have were always sorted first. Until those are measured, the fit strip is a way to read profiles faster and more consistently. It is not evidence that the people at the top are the best candidates. ### Key terms - **Fit strip**: Emoti's row of marks, one per search criterion, showing whether a candidate meets, partly meets, misses or lacks information for each criterion, with the evidence behind each mark. - **Evidence**: In Emoti, the line of profile or transcript text behind each judgment, shown next to the judgment so a person can check it. - **Must-have criterion**: A requirement a team would turn down an otherwise excellent candidate for lacking. In Emoti, must-haves count double, and missing one caps a candidate's score at 54. - **Nice-to-have criterion**: A requirement that would make a candidate stand out but whose absence wouldn't rule them out. In Emoti, nice-to-haves count once and never cap a score. - **Look-alike search**: In Emoti, a search that builds criteria from one person's profile, such as the same discipline, region, years of experience and top skills, and ranks others against them. ### Questions #### Is a candidate who misses a must-have hidden from the results? No. They stay in the list with a capped score, and the fit strip shows which must-have is missing. Removing someone from consideration is left to a recruiter. #### Which profile fields does Emoti's matcher read? Role family and seniority, location, total years of experience, listed skills, the industries of past employers, and a short list of signals such as having led a team. It doesn't read names, photos, age or gender, and it builds no criteria from schools. #### Does shortlisting or passing on someone change the ranking? Not in version 1.0. Decisions move people out of the list to review. The ranking changes only when the criteria change. #### Can two recruiters get different rankings for the same role? Yes, if their criteria differ, for example because one marked a skill as a must-have. With the same criteria, the same pool always ranks the same way. #### Why not let a language model rank candidates directly? A model's ranking is hard to reproduce and hard to explain criterion by criterion. A fixed formula over visible evidence can be checked by anyone who reads the strip. ### Sources - US Equal Employment Opportunity Commission (1978). Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4, Information on impact. Electronic Code of Federal Regulations. https://www.ecfr.gov/current/title-29/subtitle-B/chapter-XIV/part-1607/section-1607.4 ## Research paper: Grounded interview notes > Emoti's interview notes follow a template the team chooses, and every section, highlight and answer points back to the moment in the transcript it came from. A language model writes the note when one is configured; otherwise a rules engine assembles it from transcript quotes, and the app labels which one answered. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Research paper, version 1.0. For hiring managers. Canonical: https://emoti.ai/research/grounded-interview-notes ### Abstract Emoti writes interview notes in the shape of a template the team chooses, and ties each note back to the transcript: sections list the moments they draw on, answers to questions quote the transcript with times, and every time opens the transcript at that moment. When a language model is configured, it writes notes and answers under instructions to use only what the transcript says, and Emoti discards any reference that doesn't match a real turn. When no model is available, a deterministic rules engine assembles notes and answers from existing sections and transcript quotes, and the interface says which one answered. This paper describes templates, both writing paths, the Ask feature, how quotes link to moments, how people edit and restore notes, and the failure modes we know about, including references that point to the wrong turn, paraphrase that word matching misses, and transcription errors that no later step can fix. ### What does grounded mean in an interview note? A note is grounded when every statement in it can be traced to a moment in the conversation, and a reader can reach that moment in one click. Interview notes are read by people who weren't on the call: hiring managers, panel members, sometimes a reviewer months later. A fluent note that can't be checked is worse than a short one that can, because it invites a debrief to argue from memory. Research points the same way. In one study, taking notes improved how accurately interviewers recalled what candidates said but not the accuracy of their judgments, while reviewing conventional notes did improve judgment (Middendorf & Macan, 2002). US federal guidance on structured interviews asks for notes good enough to document the reasoning behind each rating, with actual examples of answers (US Office of Personnel Management, 2008). Notes are the record that judgment is checked against. Emoti's notes follow five rules: 1. Write only from the transcript. 2. Point to the moments each section draws on. 3. Say plainly when a topic wasn't discussed. 4. Label text that Emoti wrote. 5. Keep a person's edits visible and reversible. ### How do templates shape a note? A template is an ordered list of sections, each a heading plus a one-line instruction, and every note has exactly those sections in that order. Emoti includes templates for a recruiter screen, a technical interview, a portfolio review, a behavioral interview, a hiring manager intake, a debrief, a reference check, a client call, a one-page candidate brief, and a plain list of questions and answers. Teams can write their own. The recruiter screen template looks like this: | Section | Instruction Emoti follows | | --- | --- | | Why they're looking | What is prompting a move now, in their words. Note anything that would make them stay. | | Current role | Scope, team, and the most relevant work for this role. Include numbers they gave. | | Fit with the brief | Compare against the role's must-haves. Mark each as clear, partial or not discussed. | | Compensation and logistics | Salary expectations, notice period, location, visa, other processes and timing. | | Next steps | What was agreed and who owns it. | The template decides what the note pays attention to. That is useful, because it puts every candidate's evidence in the same places. It is also a responsibility: a section that asks for concerns will tend to find some. Choosing a different template for a recorded call rewrites the note under the new sections. ### How does Emoti write a note? With a model configured, a language model writes the note from the full transcript and Emoti checks its references. Without one, a rules engine rebuilds the note from quotes. Either way, the note has the template's sections in the template's order. #### With a language model Emoti sends the model the template and the transcript, with a start time on every turn, and asks for one section per template section, each listing the start times of the turns it drew on. The instructions are short: use only what the transcript says; prefer one fact per line, with specific numbers and short quotes; and if the call didn't cover a section, write one line saying it wasn't discussed. The reply must fit a fixed structure of heading, body and references. Emoti then checks the references. Any time that doesn't match the start of a real turn is dropped before the note is saved. If the model declines, returns nothing usable, or the request fails, Emoti uses the rules engine instead. Requests go through Emoti's model gateway, which runs them on the models a workspace admin assigns to each task and tries the next one in order if a model fails; the method doesn't depend on which model answers. #### Without a model The rules engine works on words. It lowercases the text, drops short and common words, and compares what is left. For each template section: 1. It counts how many words from the section's heading and instruction appear in each existing note section. If the best match shares at least two words, that text and its references are reused under the new heading. 2. If no section is close enough, it picks up to three transcript turns that share the most words with the section, puts them in time order, and quotes each with the speaker's name and time, cut at 180 characters. 3. If no turn shares any words, the section says "Not discussed in this call." When the rules engine rebuilds a note, the app says so. Some edits also need a model: a section can be shortened or turned into bullets without one, but adding detail and translating can't be done by rules, and the app says that too. #### What a grounded section looks like Here is how a "Fit with the brief" section might read for a backend role whose must-haves are six years of backend work, production Go and on-call experience. The example is invented. - Six years building backend services: clear. "Eight years, the last four on the payments platform." (3:12) - Production Go: partial. Uses Go for internal tools; the services they own are in Java. (7:45) - On call for a service they helped build: not discussed. The section ends with a "From the call" row showing 3:12 and 7:45, and each time opens the transcript at that moment. The line that says "not discussed" matters as much as the others. It tells the panel what the next interviewer should ask, instead of leaving a gap that reads as a pass. ### How does Ask answer questions about a call? Ask answers one question about one call by quoting the transcript, with a time after each point, and labels every answer with what produced it. With a model, the question and the timestamped transcript go to the model with instructions to use only the transcript, quote short phrases, put the time after each point, stay under 150 words, and, if the call didn't cover the question, say so in one sentence and suggest what to ask next time. The answer appears as it is written. Each question is answered on its own; earlier answers stay on the page but aren't sent as context. Without a model, the rules engine answers in one of four ways: - **Questions about concerns, strengths or motivation** are answered from the call's highlights of that type, when it has any, each with its title, time and quote. - **Questions about what the candidate asked** return the turns from people outside the hiring team that contain a question mark. - **Other questions** become a word search. Common words and participants' names are ignored, and a few groups of related words are expanded, so "salary" also looks for "compensation", "base" and "equity". Words match on their first five letters, so "relocate" finds "relocation". Turns are scored by shared words with a small preference for the candidate's own answers, and the best match is kept along with anything at least 60 percent as good, up to two. When a match is the interviewer's question, the reply that follows it is added. At most four turns are quoted. - **If nothing matches**, the answer says so and points to the call's first highlights. Each answer is labeled "Answered by Emoti" when a model wrote it, or "Matched from the transcript by Emoti's built-in rules" when the rules engine did. ### How does a reader check and change a note? Every time in a note, a highlight or an answer is a link to the transcript, and every section can be edited and restored. #### Links to moments A time written as minutes and seconds in square brackets becomes a small button. Pressing it switches to the transcript, scrolls to the turn that was under way at that second, and highlights it. Note sections end with a "From the call" row of times taken from their references. Highlights carry a type (strength, concern, question, logistics or motivation), a short title, a quote and a time. The transcript can be searched: matching turns stay bright, the rest are dimmed, and the number of matches is shown. Beside it, a map of who spoke when and the share of talk time on each side show the shape of the call. #### Edits and sharing A section can be edited by hand, shortened, rewritten as bullets, given more detail or translated. An edited section is marked "Edited", and "Restore what Emoti wrote" brings back the original. Summaries and answers that Emoti wrote carry a label that says so. A note can be shared by link: the shared page shows the summary, notes and highlights, while the transcript and recording stay private to the workspace. A note can also be copied in one step for pasting into an applicant tracking system. #### Recording disclosure The notetaker is set to tell people they are being recorded. With disclosure on, which is the default, the calendar invite says the call is recorded and transcribed and what the notes are used for, the notetaker joins under a visible name, and it posts a message in the call chat when recording starts. If anyone asks it to stop, it is designed to leave and delete the recording. A team can turn disclosure off, and the settings page then warns that they are responsible for telling people and, in many places, for getting their consent first. ### Limitations Grounding narrows what can go wrong. It doesn't remove it. - **Transcription errors flow into everything.** If a word is misheard or a turn is given to the wrong speaker, the note, the quotes and the answers inherit the error. Speech recognition is not equally accurate for everyone: in a study of five commercial systems, the average word error rate was 0.35 for Black speakers and 0.19 for white speakers (Koenecke et al., 2020). - **A reference can be real and still wrong.** Emoti checks that each reference points to a turn that exists. It doesn't check that the turn supports the line it is attached to. - **Quotes in model-written text are not checked word for word.** A quote can be shortened or paraphrased, and a model can in principle invent one. A wrong time opens whatever turn was under way at that second. - **Summaries can lose hedges.** "I think we cut it by about a third" can become "cut it by a third". The instruction to quote short phrases reduces this; reading the moment removes it. - **Word matching misses meaning.** The rules engine finds turns that share words with the question. It misses paraphrase, can pick a turn that mentions a word in passing, and can reuse an existing section under a heading it only loosely fits, because two shared words are enough. - **"Not discussed" means "not found".** Both paths can miss a topic that was covered in other words. - **Notes record what was said, not whether it is true.** A grounded note can faithfully quote an inaccurate claim. ### What we have not measured yet We have not yet measured: - How often each line of a model-written note is fully supported by the turns it references, by call type and by language. - How often the rules engine picks the turns a person would pick for the same question. - How transcription accuracy varies by accent, language and audio quality in real calls, and how those errors change the notes. - Whether debriefs that use grounded notes reach different decisions, or reach them with more agreement. - How well translated sections keep the meaning of the original. These are the measurements the method needs before anyone should rely on a note more than on their own reading of the transcript. ### Key terms - **Notetaker**: Software that joins a call, records and transcribes it, and drafts notes. Emoti's notetaker writes notes in the template chosen for each kind of call and, by default, tells everyone the call is recorded. - **Transcript highlight**: In Emoti, a marked moment in a call's transcript with a type (strength, concern, question, logistics or motivation), a short title, a quote and a time that opens the transcript at that moment. - **Evidence**: In Emoti, the line of profile or transcript text behind each judgment, shown next to the judgment so a person can check it. - **Debrief**: The meeting after interviews where interviewers share their ratings and evidence and the hiring team decides. It works best when every interviewer has recorded ratings independently beforehand. ### Questions #### Does Emoti make hiring decisions from interview notes? No. Notes, highlights and answers are records for people to read. Every hiring decision is made by a person. #### Can a team write its own note template? Yes. A template is a list of section headings with one instruction each. Personal templates appear beside the built-in ones, and choosing one rewrites the note for that call. #### Which languages can Emoti's notes be written in? The notetaker can write notes in the language spoken most on the call or in a chosen language, including English, Mandarin Chinese, Malay, Hindi, Japanese, German, French, Spanish, Portuguese and Polish. Translating a single section needs a language model. #### Why keep a rules engine when a model writes better notes? So every feature still works without a model, and so a team can see exactly how an answer was found. Rules answers only quote the transcript, and they are always labeled as coming from rules. ### Sources - Middendorf, C. H., & Macan, T. H. (2002). Note-taking in the employment interview: Effects on recall and judgments. Journal of Applied Psychology, 87(2), 293–303. https://doi.org/10.1037/0021-9010.87.2.293 - US Office of Personnel Management (2008). Structured Interviews: A Practical Guide. US Office of Personnel Management. https://www.opm.gov/policy-data-oversight/assessment-and-selection/structured-interviews/guide.pdf - Koenecke, A., Nam, A., Lake, E., Nudell, J., Quartey, M., Mengesha, Z., Toups, C., Rickford, J. R., Jurafsky, D., & Goel, S. (2020). Racial disparities in automated speech recognition. Proceedings of the National Academy of Sciences, 117(14), 7684–7689. https://doi.org/10.1073/pnas.1915768117 ## Research paper: Personal outreach with a person in the loop > Emoti drafts one personal first line for each candidate from facts in their own professional profile, and nothing is sent until a recruiter approves the line and the sequence is active. Emails go out on weekdays inside a sending window and under a daily limit, and any reply stops the sequence for that person. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Research paper, version 1.0. For recruiters. Canonical: https://emoti.ai/research/outreach-with-approval ### Abstract Emoti drafts a personal first line for every person in an outreach sequence from facts in that person's own professional profile, and nothing is sent until a recruiter approves the line and starts the sequence. This paper describes how the first line is composed in version 1.0, which facts it may and may not use, how approval works, how sequences are timed and paced, and how replies and bounces stop a sequence. It sets out the tests we use to keep a message personal without making it intrusive, and it names what version 1.0 does not do yet, including a one-click way to opt out of all future messages and a workspace-wide do-not-contact list. ### What is a first line, and how does Emoti draft it? A first line is the opening sentence of the first message to a candidate. It is the only part of the message written for one person, and its job is to show that someone looked at their work before writing. In version 1.0 Emoti drafts it with a deterministic composer, not a language model, from evidence in the person's profile. The rest of the message is shared by everyone in the sequence, with placeholders for details such as the person's first name, their company and the sender's signature. One placeholder is Emoti's own: the personal first line. Wherever a message is shown in Emoti, that placeholder is highlighted differently from the others, so it is always clear which words Emoti will write. A first message in Emoti's example sequences is built like this, with the personal line straight after the greeting: > Hi [first name], [personal first line] I'm [sender], and I recruit for [team]. [Two or three sentences about the work and the role.] Would you be open to a 20-minute call next week? If you'd rather read first, I can send the brief. [Signature] #### The four angles The composer looks for four kinds of evidence, writes a line for each kind it finds, and picks one. | Angle | Evidence used | Shape of the line | | --- | --- | --- | | Work | The first clause of their current role's description that starts with a verb | "You own the pricing models used in every quote at [company], and I'd love to hear how you'd approach [what the role is about]." | | Skills | Two skills on their profile that the role's brief also names, and the year they joined their current company | "[Skill] and [skill] are a big part of how our [team] team works, and you've been using both at [company] since [year]." | | Industry | Past work in the hiring company's industry or a closely related one | A sentence about why that background carries over | | Signals | One of a short list: conference speaker, open-source maintainer, has led a team, mentor, startup experience, former founder | A sentence about what that habit says about their work | Two rules sit on top. If the profile says the person was recently promoted, the line congratulates them on the new title and acknowledges that a note like this "lands at an odd moment". If no angle applies, a fallback line names their years in their discipline and their current company. When several angles apply, one is chosen by a stable hash of the person and the role. The same person and role always get the same line. That keeps the server and the browser in agreement, and it means the line a recruiter approves is the line that goes out. #### Why a composer and not a model Every clause the composer writes can be traced to a field in the profile, and it can't invent an employer, a project or a number. That is why version 1.0 uses it. The cost is that the lines can sound formulaic, and their grammar depends on a few simple rules for turning "Owns pricing models" into "you own pricing models". Any move to a language model would need the same guarantees: every claim traceable to a field in the profile, and a person's approval before anything is sent. ### What may a first line use, and what is off limits? A first line may use work facts the person published on a professional profile. It never uses their personal life, and it never uses private status. In version 1.0 a first line can draw on the description of the person's current role, their listed skills, the industries of their past employers, the year they joined their current company, and the signals listed above. It never uses their location, education, photo, age or family, anything outside their professional profile, or whether they are open to work. That last one deserves a note. Whether someone is looking for a job is private context, even when a profile shows it. A recruiter may use it to decide whom to contact, but it never goes into a message. The line between personal and intrusive is easier to feel than to define. We use four tests: 1. **Is it professional?** It is about their work, not their life. 2. **Is it public and meant to be seen?** They published it about their work; it wasn't found by combining sources. 3. **Is it recent and still true?** A stale fact reads as careless. 4. **Is it relevant to the role?** If it doesn't explain why you are writing, leave it out. Some invented examples show the difference: | Personal | Intrusive | | --- | --- | | "Your talk on retry storms made the case for idempotency keys better than most documentation does." | "I saw you were at the conference with your family last week." | | "You've maintained the scheduling library for four years." | "I noticed your profile now says you're open to work." | | "You moved from consulting into building the product itself." | "You've been in your role for three years, so you're probably due a move." | The promotion line shows how a fact can pass every test and still be unwelcome. Some people are glad to be congratulated, and some don't want to hear from recruiters weeks into a new title. That is one reason a person approves each line. In the EU, when personal data is not collected from the person directly, the GDPR requires telling them, among other things, where the data came from and how it will be used, and if the data is used to contact them, doing so at the latest at the first communication (Article 14). Emoti doesn't add that notice to messages in version 1.0, so teams recruiting in the EU need to include it themselves. This paper is not legal advice. ### How does approval work? Every person added to a sequence starts as scheduled, with nothing sent. Their first line waits in a list until a recruiter approves it, and messages go out only once the line is approved and the sequence is active. The list shows each person's name and drafted line with an Approve button. When more than one line is waiting, an "Approve all" button appears. Lines already sent are kept for reference. New sequences start as drafts, and pausing a sequence stops everything that hasn't been sent until it is resumed. Text Emoti wrote is labeled as such. A recruiter can edit a line before approving it, or leave it unapproved so that person isn't contacted. Each approval is saved with the approver's name and the time, and with the edited text when the line was changed. "Approve all" saves time, and it also makes it easy to stop reading. Approval protects candidates only if the approver reads. Before approving, a reviewer should check five things: 1. The fact is true and current. If in doubt, open the profile. 2. It is about their work, not their life. 3. It doesn't suggest you have been watching them. 4. It explains why this role, and not some other role, made you write. 5. The rest of the message gives the role, the location and the pay range. ### How are sequences timed and paced? A sequence is an ordered list of steps on set days: emails, LinkedIn messages, call tasks and waits. Emails go out on weekdays, inside a sending window in each recipient's own time zone, under a daily limit for each mailbox. The defaults are a sending window of 8:30 a.m. to 5:30 p.m. and a limit of 60 emails a day per mailbox, shared across every sequence that uses it. A team can set the limit anywhere from 1 to 500, and anything over it waits for the next sending window. The settings page gives the reason: steady volumes keep a mailbox out of spam folders. When shortlisted people are added to a new sequence, Emoti drafts three messages for the role: an email on day 1, a follow-up email on day 5 after a wait, and a LinkedIn message on day 9. Call steps, where a sequence has them, are tasks for the recruiter, with a note on what to say. Each person's progress is tracked as a status: | Status | What it means | Does the sequence continue? | | --- | --- | --- | | Scheduled | Added, but nothing has been sent yet | Not started | | Active | Moving through the steps | Yes | | Replied, Interested or Not interested | They wrote back | No, their reply stopped it | | Bounced | An email couldn't be delivered | No | | Finished | Every step has gone out | No | ### How can a person opt out? In version 1.0, by replying. Any reply stops the sequence for that person, whether it says yes, not now or no. Emoti's example sequences also end with a plain last note that says it is the last one. A reply that says "not now" and one that says "please stop" look the same to a sequence: both stop it. The difference matters later, when a colleague considers contacting the same person about another role, so the answer needs to be recorded in the person's own words as well as in a status. There is no one-click unsubscribe link and no workspace-wide do-not-contact list yet. A person who says no in one sequence could later be added to another by a colleague who doesn't know. This is the most important gap in the outreach design. Until it is closed, we recommend that every first message include a plain sentence about stopping, and that teams record those answers where everyone who adds people to sequences can see them. Emoti's sample sequences now end their first email with one: "Not for you? Reply and say so, and I won't write again." ### Limitations - **Lines can sound templated.** A reader who gets several recruiter messages may recognize the shapes. - **The grammar rules are simple.** Turning a clause from a profile into a sentence about "you" can produce an awkward phrase. - **The skills line makes an inference.** It says the person has used both skills since they joined their current company, which the profile doesn't state. - **Profiles go stale.** A promotion, a team or a project mentioned in a line may be out of date by the time it is sent. - **The description of the role is written once** and repeated in every line that uses it. - **"Approve all" invites skimming**, because approving many lines at once makes it easy to stop reading. - **Opt-outs don't carry across sequences**, and the EU notice isn't added automatically. - **Pacing is per mailbox.** Nothing limits how many people at one company are contacted on the same day. - **Lines are written in English only.** The composer has no other languages in version 1.0. ### What we have not measured yet We have not yet measured: - Whether personal first lines change reply rates. We make no claim that they do. - How often recruiters decline lines, and which angles they decline most. - How candidates perceive the lines: personal, generic or intrusive. - How often the facts in a line are out of date by the time it is sent. - How sending windows and daily limits affect whether emails are delivered. - Whether reviewers catch the composer's mistakes, such as a stale fact or an awkward clause, before they approve a line. ### Key terms - **First line**: In outreach, the opening sentence of the first message, written for one person from their own work. Emoti drafts one per person and a recruiter approves it before anything is sent. - **Outreach sequence**: A planned series of messages and tasks, such as emails, LinkedIn messages, calls and waits, sent to a candidate over days or weeks until they reply or the sequence ends. - **Passive candidate**: A person who isn't actively looking for a new job but might consider the right one. Sourcing and outreach are how teams reach them. - **Sourcing**: Finding and identifying people who could fit a role, especially people who haven't applied, through searches, referrals and research. ### Questions #### Does Emoti send outreach without a person's approval? No. A person's first line must be approved and the sequence must be active before their first message goes out. Pausing a sequence stops everything that hasn't been sent. #### Why does the same candidate always get the same first line? The line is picked by a stable rule, so the line a recruiter reviews is the line that is sent, on any device. A different role can produce a different line for the same person. #### How many emails can a mailbox send in a day? Up to its daily limit, which is 60 by default and can be set from 1 to 500. The limit is shared across every sequence that sends from that mailbox. #### What happens when someone replies "not now"? Their reply stops the sequence. In version 1.0 Emoti doesn't schedule a later check-in, so recruiters note the date themselves. ### Sources - European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Article 14. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj ## Article: How to write a hiring brief an AI agent can act on > A hiring brief an AI agent can act on names two to four must-haves, a few nice-to-haves, the evidence that would prove each one, and the location, level, pay range and deal-breakers, all in plain words. The agent turns the checkable parts into criteria it can test against each profile, so a specific brief produces a ranking you can audit and a vague one produces almost nothing. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For recruiters. Canonical: https://emoti.ai/articles/writing-a-hiring-brief ### What is a hiring brief? A hiring brief is a short, plain-language description of who you want to hire and how you will recognize them, written before anyone starts searching. It is different from a job description, which describes the job to candidates, and a job post, which advertises it. The brief is for the people and tools doing the search. It usually comes out of an intake meeting between the recruiter and the hiring manager, and it should be specific enough that two recruiters reading it would bring back similar shortlists. An AI agent raises the stakes. A colleague can ask what you meant by "strong". An agent turns your words into criteria and ranks people against them, so every vague word becomes a vague criterion, or no criterion at all. ### What should a brief include? Seven things: the role and level, a few must-haves, some nice-to-haves, the evidence you would accept for each, the location and work pattern, the pay range, and the deal-breakers. | Part | What to write | Example | | --- | --- | --- | | Role and level | The title people actually hold, and the level | Senior backend engineer, individual contributor | | Must-haves | Two to four requirements you would turn a strong candidate down for lacking | 6+ years building backend services | | Nice-to-haves | Things that would make someone stand out | Kafka or another event streaming system | | Evidence | What you would accept as proof of each requirement | Owned a service in production, including on-call | | Location | City, region or remote, and office days | Berlin with two office days a week, or remote within Europe | | Pay | The base range, and anything else in the package | €95,000 to €115,000 base | | Deal-breakers | Conditions that end the conversation | Can't join the on-call rotation | Write the pay range down even if the agent can't use it. You will need it in the first message and the first call, and some places require it in job postings. In California, for example, employers with 15 or more employees must include the pay scale in any job posting. ### How do you separate must-haves from nice-to-haves? A must-have is something you would turn down an otherwise excellent candidate for lacking. Everything else is a nice-to-have. The test is simple. Picture the best person you could hire for this role, then take the requirement away. Would you still hire them? If yes, it is a nice-to-have. Aim for two to four must-haves. Each extra one shrinks the pool, and requirements that sound essential in a meeting, such as a specific framework or industry, often turn out to be preferences. Must-haves should also be checkable. "Excellent communicator" can't be checked from a profile. "Has written design documents that other teams used" can be checked in a screen. ### What evidence would you accept for each requirement? For every requirement, write down what you would see in a profile or hear in a screen that proves it, because most requirements can be shown in more than one way. "Five years of Python" might mean five years of job titles that mention Python, or open-source work, or a long career in which Python was one tool among several. Decide in advance what counts. For example: - **Owned a production service:** a role where they ran a service, were on call for it, or led its incident reviews. - **Go:** production Go in a recent role, or public Go projects that other people use. - **Worked closely with sales:** built tools for a sales team, or sold before moving into engineering. Two kinds of evidence are worth separating. What a profile can show helps with sourcing. What only a conversation can show belongs in the screen. Structured interviewing starts from the same place, by basing questions on an analysis of what the job requires (Campion, Palmer & Campion, 1997). Avoid proxies. School names, famous employers and years in a role stand in for skills without measuring them, and they narrow the pool in ways that have little to do with the work. ### What does a weak brief look like next to a strong one? A weak brief describes a kind of person. A strong brief describes the work and how you will recognize someone who can do it. > We need a rockstar senior engineer who can hit the ground running. Ideally ex-big tech, top school, culture fit. Competitive pay. Remote-ish. > Senior backend engineer for our payments team. Berlin, with two office days a week, or remote within Europe. Must have: 6+ years building backend services; production Go or Rust; has been on call for a service they helped build. Nice to have: Kafka or another event streaming system; payments or marketplace background. Evidence we'd accept: a role where they owned a service in production; public Go or Rust work counts. Level: senior individual contributor, no direct reports. Pay: €95,000 to €115,000 base. Deal-breaker: can't join the on-call rotation. When we ran both briefs through Emoti's built-in rules, the weak one produced a single criterion: "Open to remote". "Rockstar", "hit the ground running" and "culture fit" can't be checked, "senior engineer" doesn't say what kind of engineer, and "competitive pay" says nothing at all. The strong brief produced seven criteria: | Criterion | Must-have or nice-to-have | | --- | --- | | Senior backend engineer | Must-have | | Based in Berlin or Europe, or remote | Nice-to-have, because remote work is allowed | | 6+ years of experience | Must-have | | Go | Nice-to-have | | Rust | Nice-to-have | | Kafka | Nice-to-have | | Experience in marketplace or payments | Nice-to-have | Two things in that list need a person's attention. Go and Rust came in as nice-to-haves even though the brief calls them must-haves, because skills start as nice-to-haves by default. And "Go or Rust" became two separate criteria, because Emoti can't yet express "either one". Marking both as must-haves would require both. The fix is to leave them as nice-to-haves and read the evidence, or to pick the one that matters more. ### How does Emoti read a brief into criteria? Emoti turns the brief into a short list of criteria, marks each as a must-have or a nice-to-have, and shows them as chips you can change before you trust the ranking. Criteria cover the role, location, years of experience, skills, industry and a few signals, such as having led a team. The role, a stated location and the minimum years start as must-haves; skills and industry start as nice-to-haves; and a brief that allows remote work makes location a nice-to-have. When a language model is configured, it reads the brief. Otherwise Emoti's built-in rules do, and the page says which one did. Every candidate then gets a fit strip: one mark per criterion, showing met, partly met, missing or not enough information, with the line of profile evidence behind each mark. Missing a must-have doesn't hide anyone. It caps their score at 54 out of 100, which keeps them out of the strong and good bands, and the strip shows which must-have they miss. A coverage panel shows how many people in the pool meet each criterion and warns when a must-have is so rare that it is shrinking the search. Some parts of a brief aren't read by the matcher in version 1.0: the pay range, the deal-breakers and your evidence notes. They are for the people reviewing the shortlist and for the questions in the first screen. The full method, including its limits, is in [Evidence-first candidate matching](https://emoti.ai/research/evidence-first-matching). ### Which mistakes make a brief hard to act on? Most weak briefs fail in one of a few ways: - **Too many must-haves.** A list of eight must-haves is often three must-haves and five preferences. - **Adjectives without a test.** "Strong", "excellent" and "senior-minded" each need a sentence about what you would actually see. - **Proxies for skill.** Schools, famous employers and years of experience stand in for the thing you want. - **Hidden deal-breakers.** A relocation or travel requirement that appears at the offer stage wastes everyone's time. - **Two roles in one.** "Backend engineer who can also lead design" is usually two searches. - **A pasted job description.** It is written to attract applicants, not to rank them, and it mixes requirements with marketing. - **Words that describe a kind of person.** Phrases such as "young and hungry" or "digital native" describe who someone is, not what they can do. Leave them out. ### Key terms - **Hiring brief**: A short, plain-language description of who a team wants to hire and how it will recognize them: must-haves, nice-to-haves, evidence, location, level, pay range and deal-breakers. - **Must-have criterion**: A requirement a team would turn down an otherwise excellent candidate for lacking. In Emoti, must-haves count double, and missing one caps a candidate's score at 54. - **Nice-to-have criterion**: A requirement that would make a candidate stand out but whose absence wouldn't rule them out. In Emoti, nice-to-haves count once and never cap a score. - **Evidence**: In Emoti, the line of profile or transcript text behind each judgment, shown next to the judgment so a person can check it. - **Fit strip**: Emoti's row of marks, one per search criterion, showing whether a candidate meets, partly meets, misses or lacks information for each criterion, with the evidence behind each mark. - **Intake meeting**: The meeting where a recruiter and hiring manager agree on a role before the search starts: what the person will do, the must-haves, where to look and the interview process. ### Questions #### How long should a hiring brief be? Long enough to state each requirement and the evidence for it, and short enough to read in a minute. Most briefs fit in one paragraph or a short list. #### Who should write the hiring brief? The recruiter and the hiring manager together, usually in an intake meeting. The hiring manager owns the must-haves; the recruiter makes sure each one can be checked. #### Can I change the brief after the search has started? Yes. In Emoti, editing the brief rebuilds the criteria and keeps your shortlist decisions. You can also switch any single criterion between must-have and nice-to-have. #### What if almost nobody meets every must-have? Find the rarest must-have and ask whether you would really turn down a strong candidate without it. Emoti's coverage panel flags any must-have that fewer than 8 percent of the pool meet. ### Sources - Campion, M. A., Palmer, D. K., & Campion, J. E. (1997). A review of structure in the selection interview. Personnel Psychology, 50(3), 655–702. https://doi.org/10.1111/j.1744-6570.1997.tb00709.x - Labor Commissioner's Office California Equal Pay Act and Pay Transparency Law (Labor Code sections 432.3 and 1197.5): Frequently Asked Questions. California Department of Industrial Relations. https://www.dir.ca.gov/dlse/California_Equal_Pay_Act.htm ## Article: Interview notes that hold up in a debrief > Interview notes hold up in a debrief when they record what the candidate said and did rather than how the interviewer felt, with short quotes and times, organized by the competencies on the rubric. Write them up straight after the interview, before talking to anyone, and tell candidates when a call is recorded, because consent rules differ from place to place. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For hiring managers. Canonical: https://emoti.ai/articles/interview-notes-that-hold-up ### What makes a note useful in a debrief? A useful note lets someone who wasn't in the room see the evidence behind each rating, without having to trust the writer's memory or mood. Notes seem to help memory more than judgment. In one study, interviewers who took notes recalled what candidates said more accurately, but their judgments were no more accurate; reviewing conventional notes before judging did help (Middendorf & Macan, 2002). So the value of a note lies in what it lets you and the panel check later. The US Office of Personnel Management's guide to structured interviews asks for notes good enough to document the interviewer's reasoning for each rating, including actual examples of answers (US Office of Personnel Management, 2008). A note that holds up answers five questions for each competency: what you asked, what the candidate said or did, the words that matter most, the rating you gave and the anchor it matched, and what you didn't get to. If a reader can't find those, the note will be argued about rather than used. ### How do you write evidence instead of impressions? Write down what the candidate said or did, then, separately, what you concluded from it. | Impression | Evidence | | --- | --- | | Great communicator | Explained the retry design in two minutes without jargon, then answered the follow-up on duplicate events with a concrete example | | Not senior enough | Described decisions their lead made; asked what they would have changed, said "I'd have to think about it" | | Seemed unsure about the numbers | Gave two different team sizes, six and then nine | | Culture fit | Leave it out, or write the specific behavior you mean | Impressions aren't useless. They are conclusions, and the debrief needs the observations they rest on. If you can't find an observation behind an impression, that is worth knowing too. Some things should never be in notes at all: comments on appearance, accent, age, family, health, or anything else unrelated to the job. Write every note as if the candidate might one day read it. ### Why include short quotes and times? A short quote with a time lets the panel check the context instead of arguing about what was said. Quote the key words, not the whole answer. "We cut the queue from 40 minutes to 5 by batching writes" is worth more than "talked about performance work". If the call was recorded, add the minute. When two interviewers remember an answer differently, a time turns an argument into a 30-second replay. ### How should notes be organized? By the rubric: one section per competency, each with the evidence, the rating and anything to probe next. | Competency | Evidence | Rating | To probe next | | --- | --- | --- | --- | | Owns production systems | On call for the billing service; led the review after a double-charge incident | 4 of 4 | How they chose which fixes came first | | Works across teams | "I set up a weekly sync with finance" | 2 of 4 | What changed because of it | Notes organized by competency make scoring faster and make gaps visible. If a section is empty, the interview didn't cover that competency, and the panel should know that before it decides. For behavioral questions, capture three parts in the candidate's own terms: the situation, what they did, and what happened. When one part is missing, write that down too ("no result given"), because a story without an outcome is weaker evidence than it sounds in the moment. ### When should you write the notes up? Straight after the interview, before you talk to anyone else about the candidate. Take brief notes during the call, then spend a few minutes turning them into evidence under each competency and recording your ratings. The OPM guide describes the same sequence: interviewers review their notes and rate the candidate immediately afterward, and panel members form independent evaluations before any discussion. Talking first can let the strongest opinion in the hallway rewrite everyone's memory of the interview. A five-minute routine helps: 1. Fill in the evidence under each competency while it is fresh. 2. Mark the one or two quotes that matter most, with their times. 3. Rate each competency against its anchors, not against the other candidates. 4. Note what you didn't cover, so the next interviewer can. 5. Submit your notes and ratings before you read anyone else's. ### How should you tell candidates a call is recorded? Tell them before the call and again at the start, say why it is recorded and who will see it, and give them a way to say no. Rules on consent differ by place. In the US, federal law allows a recording when one party to the conversation consents (18 U.S.C. § 2511(2)(d)), but some states require the consent of every party to a confidential conversation. California is one example (California Penal Code section 632). Other countries have their own rules, including data protection law. The simplest policy that works everywhere is to tell everyone, every time. This article is not legal advice. A sentence at the start is enough: "I'd like to record this call so I can listen properly instead of typing, and so the hiring team can check what was said. Only the people involved in this decision will see it. Is that all right with you?" If the candidate says no, don't record. Take notes by hand, and make sure their answer doesn't count against them. ### Where do AI notetakers fit? An AI notetaker records and transcribes the call and drafts notes in your template, so you can listen instead of typing. You still own the ratings, and you should check the draft against the transcript before you rely on it. Emoti's notetaker writes notes in the template chosen for each kind of call, links each section to the moments it draws on, and answers questions about the call by quoting the transcript with times. When no AI model is available, a rules engine assembles notes and answers from transcript quotes, and the app says which one answered. Recording disclosure is on by default: the calendar invite says the call is recorded, the notetaker joins under a visible name, and it posts a message in the call chat when recording starts. Notetakers fail in ways worth knowing. Speech recognition makes more errors for some speakers than others; one study of five commercial systems found an average word error rate of 0.35 for Black speakers and 0.19 for white speakers (Koenecke et al., 2020). A summary can also drop a hedge such as "I think" or "roughly". Read the moment before you quote it in a debrief. Before a draft goes to the debrief, check four things: the quotes you rely on match the transcript, the speakers are labeled correctly, nothing in the draft is an impression you wouldn't have written yourself, and any section that says a topic wasn't discussed really wasn't. How Emoti's notes are built, and where they can go wrong, is in [Grounded interview notes](https://emoti.ai/research/grounded-interview-notes). ### Key terms - **Debrief**: The meeting after interviews where interviewers share their ratings and evidence and the hiring team decides. It works best when every interviewer has recorded ratings independently beforehand. - **Scorecard**: The record of one interviewer's or one screen's ratings for a candidate, competency by competency or question by question, with the evidence for each rating. - **Rubric**: A written guide for rating answers: for each question or competency, what a strong, middling and weak answer contains. - **Notetaker**: Software that joins a call, records and transcribes it, and drafts notes. Emoti's notetaker writes notes in the template chosen for each kind of call and, by default, tells everyone the call is recorded. - **Transcript highlight**: In Emoti, a marked moment in a call's transcript with a type (strength, concern, question, logistics or motivation), a short title, a quote and a time that opens the transcript at that moment. ### Questions #### Should I take notes during the interview? Yes, brief ones. Write key words and short quotes during the call and expand them straight afterward. The OPM guide also suggests telling the candidate at the start that you will take notes. #### Can candidates see interview notes about them? Sometimes. In the EU, people have the right to a copy of the personal data an organization holds about them under Article 15 of the GDPR. Write every note as if the candidate may read it. #### How long should interview notes be kept? As long as your retention policy says, and no longer. In the EU, the GDPR requires personal data to be kept no longer than is necessary for its purpose. #### Should I read other interviewers' notes before the debrief? Only after you have recorded your own ratings. Reading first makes it hard to tell what you saw from what someone else wrote. ### Sources - Middendorf, C. H., & Macan, T. H. (2002). Note-taking in the employment interview: Effects on recall and judgments. Journal of Applied Psychology, 87(2), 293–303. https://doi.org/10.1037/0021-9010.87.2.293 - US Office of Personnel Management (2008). Structured Interviews: A Practical Guide. US Office of Personnel Management. https://www.opm.gov/policy-data-oversight/assessment-and-selection/structured-interviews/guide.pdf - United States Congress 18 U.S.C. § 2511, Interception and disclosure of wire, oral, or electronic communications prohibited. United States Code, via govinfo.gov. https://www.govinfo.gov/content/pkg/USCODE-2021-title18/html/USCODE-2021-title18-partI-chap119-sec2511.htm - California Legislature California Penal Code section 632. California Legislative Information. https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=PEN§ionNum=632 - Koenecke, A., Nam, A., Lake, E., Nudell, J., Quartey, M., Mengesha, Z., Toups, C., Rickford, J. R., Jurafsky, D., & Goel, S. (2020). Racial disparities in automated speech recognition. Proceedings of the National Academy of Sciences, 117(14), 7684–7689. https://doi.org/10.1073/pnas.1915768117 - European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Articles 5 and 15. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj ## Article: Structured interviews, a practical guide > A structured interview asks every candidate the same job-related questions and rates each answer against a written, anchored scale, with interviewers scoring independently before they compare notes. Research finds structured interviews predict job performance better than unstructured conversations, and a 2022 re-analysis ranked them first among common selection methods, though results vary widely between settings. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For hiring managers. Canonical: https://emoti.ai/articles/structured-interviews ### What makes an interview structured? An interview is structured when every candidate is asked the same job-related questions and every answer is rated against the same written standard. The benefit depends on doing both consistently. A widely used framework, from Campion, Palmer and Campion (1997), lists 15 components of structure. A later review by Levashina, Hartwell, Morgeson and Campion (2014) groups them by whether they shape the content of the interview or the way answers are evaluated: | Content of the interview | Evaluation of answers | | --- | --- | | Questions based on a job analysis | Rating each answer, or using several rating scales | | The same questions for every candidate | Anchored rating scales | | Limited prompting and follow-up | Detailed notes | | Better types of questions | Several interviewers | | Longer interviews or more questions | The same interviewers for every candidate | | Control of other information, such as test scores | No discussion of candidates between interviews | | Candidates' own questions held until the end | Interviewer training | | | Combining ratings statistically rather than by judgment | You don't need all 15. In the studies Levashina and colleagues reviewed, the components used most often were job analysis, the same questions, better question types, anchored rating scales, rating each question and training. ### Why do structured interviews predict job performance better? Because they remove much of the variation in what gets asked and how answers are judged, so the ratings say more about the candidate and less about the interviewer. Researchers measure this as validity: the correlation between interview ratings and later ratings of job performance, corrected for known measurement problems. In a widely cited summary, Schmidt and Hunter (1998) reported a validity of .51 for structured interviews and .38 for unstructured ones. In 2022, Sackett, Zhang, Berry and Lievens re-analyzed the evidence. They argued that an adjustment called the correction for range restriction, which accounts for the fact that validation studies mostly include people who were hired, had often been applied in ways that overstated validity. Their revised estimates were lower for most methods, and the order changed: | Method | Schmidt and Hunter (1998) | Sackett et al. (2022) | | --- | --- | --- | | Structured interviews | .51 | .42 | | Job knowledge tests | .48 | .40 | | Empirically keyed biodata | .35 | .38 | | Work sample tests | .54 | .33 | | General cognitive ability tests | .51 | .31 | | Unstructured interviews | .38 | .19 | Structured interviews came out on top, and the gap between structured and unstructured interviews widened. The authors also stressed variation: the estimate for structured interviews had a standard deviation of .19 across studies, and a lower 80 percent credibility value of .18 (Sackett et al., 2023). An average of .42 hides interviews that work far better and far worse. Structure is not a guarantee, and a badly built structured interview can still predict poorly. Structure also appears to help with fairness. In Levashina and colleagues' meta-analysis of studies published after 1996, differences in structured interview ratings by race, gender and disability were mostly negligible, though the authors cautioned that the number of studies was small. ### How do you build a rubric with anchored ratings? Write the competencies first, then the questions, then a description of what each rating sounds like for each question. The order matters: 1. **Name three to five competencies** the job needs, taken from the hiring brief or a short job analysis. 2. **Write one or two questions per competency.** Past-behavior questions ask what someone did ("Tell me about a time you…"); situational questions ask what they would do. Researchers commonly recommend using both (Levashina et al., 2014). 3. **Anchor every point on the scale.** For each question, describe what a weak, mixed, good and strong answer contains, using the kinds of things candidates actually say. 4. **Plan your follow-ups.** Decide in advance which neutral probes you will use, and use them the same way with everyone. 5. **Try the questions on a colleague.** Fix questions that confuse and anchors that don't separate answers. 6. **Train the interviewers** on the questions, the anchors and how to take notes. Here is an anchored scale for one question: "Tell me about a time new information forced you to change a plan you'd committed to." | Rating | What the answer contains | | --- | --- | | 1, weak | No specific example, or a change that was forced on them with no decision of their own | | 2, mixed | A real example, but vague about what they decided or why | | 3, good | A clear example, the information that changed, and the decision they made | | 4, strong | All of that, plus how they told the people affected and what they would do differently | Anchors make a measurable difference. In one meta-analysis of past-behavior interviews summarized by Levashina and colleagues, interviews with anchored rating scales had higher validity (.35 against .26) and higher agreement between raters (.77 against .73). Another study they review found that a five-point scale with an anchor at every point made ratings more resistant to disability bias than a scale anchored only at its ends. ### Why should interviewers score before the debrief? Because once opinions are shared, the first and most confident voices can shape everyone else's memory of the interview. Independent scores preserve what each interviewer actually saw. The US Office of Personnel Management's guide to structured interviews describes the sequence: each panel member reviews their notes and rates the candidate right after the interview, forming an independent evaluation "without discussion with other panel members", and only then do they compare notes and reach a consensus (US Office of Personnel Management, 2008). Campion and colleagues list no discussion of candidates between interviews as a component of structure in its own right. A simple debrief follows from this: 1. Every interviewer submits scores and notes before the meeting. 2. Scores are revealed at the same time. 3. The panel discusses the largest disagreements first, using the notes. 4. The hiring manager records the decision and the reasons for it. Rating as you go may also help. An exploratory analysis by Levashina and colleagues suggests that rating after each question, rather than once at the end, might reduce differences between groups. ### What are the most common mistakes? - **Questions that drift.** If the questions change from candidate to candidate, the ratings can't be compared. - **One overall impression.** Scoring the whole interview at the end invites the halo of a good first five minutes. - **Anchors at the ends only.** "Strong" and "weak" leave the middle of the scale to taste. - **Unplanned probing.** Follow-ups that differ by candidate undo the structure. Levashina and colleagues recommend planned, neutral probes, used equally, and only when an answer is clearly incomplete or off track or the candidate is shy. - **Leaking other information.** Resumes, test scores and other interviewers' opinions can tilt ratings. Keep them out of the room until scoring is done. - **Talking between interviews.** A hallway comment becomes a frame for the next interviewer. - **No training.** A good rubric in the hands of untrained interviewers is still an unstructured interview. ### Where does AI help, and where doesn't it? AI helps by keeping the structure in place: asking the same questions, capturing what was said, and filing evidence under the right competency. Ratings and decisions should stay with trained people. In Emoti, interview notes follow a template with one section per topic, and every point links back to its moment in the transcript, which makes it easier to score from evidence rather than memory. Emoti's first-round voice screens ask every candidate the same questions and rate each answer against a rubric the team writes, and a recruiter makes every decision. Both have limits that we describe openly in [Grounded interview notes](https://emoti.ai/research/grounded-interview-notes) and [Designing first-round voice screens candidates can trust](https://emoti.ai/research/trustworthy-voice-screens). ### Key terms - **Structured interview**: An interview in which every candidate is asked the same job-related questions and every answer is rated against the same written standard. - **Rubric**: A written guide for rating answers: for each question or competency, what a strong, middling and weak answer contains. - **Anchored rating scale**: A rating scale in which each point is described by an example or description of the answer that earns it, so interviewers rate answers against the same standard. - **Competency**: A skill, behavior or kind of knowledge the job requires, defined clearly enough to be assessed, such as owning production systems or negotiating contracts. - **Debrief**: The meeting after interviews where interviewers share their ratings and evidence and the hiring team decides. It works best when every interviewer has recorded ratings independently beforehand. ### Questions #### Do structured interviews feel robotic to candidates? They don't have to. You can open with introductions and close with time for the candidate's questions; the structure applies to the scored questions in between. #### How many questions should a structured interview have? Enough to cover each competency with one or two questions. Campion and colleagues count longer interviews and more questions as a component of structure, within what a candidate can reasonably sit through. #### Should I use behavioral or situational questions? Both. Research finds good validity for both types for most jobs, and reviewers commonly recommend using them together because they tend to measure different things. #### Does a structured interview remove bias? It reduces some sources of bias but doesn't remove them. Check your outcomes, for example pass rates by group at each stage, as well as your process. ### Sources - Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124(2), 262–274. https://doi.org/10.1037/0033-2909.124.2.262 - Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range. Journal of Applied Psychology, 107(11), 2040–2068. https://doi.org/10.1037/apl0000994 - Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2023). Revisiting the design of selection systems in light of new findings regarding the validity of widely used predictors. Industrial and Organizational Psychology, 16(3), 283–300. https://doi.org/10.1017/iop.2023.24 - Campion, M. A., Palmer, D. K., & Campion, J. E. (1997). A review of structure in the selection interview. Personnel Psychology, 50(3), 655–702. https://doi.org/10.1111/j.1744-6570.1997.tb00709.x - Levashina, J., Hartwell, C. J., Morgeson, F. P., & Campion, M. A. (2014). The structured employment interview: Narrative and quantitative review of the research literature. Personnel Psychology, 67(1), 241–293. https://doi.org/10.1111/peps.12052 - US Office of Personnel Management (2008). Structured Interviews: A Practical Guide. US Office of Personnel Management. https://www.opm.gov/policy-data-oversight/assessment-and-selection/structured-interviews/guide.pdf ## Article: Taking an AI voice screen: what candidates should know > An AI voice screen is a short first-round interview in which an AI interviewer asks set questions by voice and your answers are rated against the employer's rubric; who makes the decision depends on the employer, so ask. Prepare as you would for any structured interview, ask for an adjustment or a human alternative before you start if you need one, and ask how long the recording is kept. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For job seekers. Canonical: https://emoti.ai/articles/ai-voice-screens-for-candidates ### What is an AI voice screen? An AI voice screen is a first-round interview in which a computer voice asks a set list of questions and records your spoken answers. You usually take it from a link, at a time you choose. It replaces, or comes before, the first phone call with a recruiter. The questions are usually about your experience, why you want the role, and practical details such as when you could start. The interviewer may ask a short follow-up if an answer is brief. The invitation or the first page should tell you how long it takes, and it is fair to ask if it doesn't. ### What is scored, and who decides? Usually the content of your answers, rated against a rubric the employer wrote. Who makes the decision depends on the employer, so ask. A rubric describes what a strong answer and a weak answer to each question contain. In Emoti's screens, each answer is rated on a four-point scale against that rubric, with a one-line summary, and a recruiter reads the results and decides whether you move on. Emoti doesn't reject anyone on its own. The camera, if it is used at all, never affects the score, and the screen doesn't try to read your emotions from your face or your voice. Other tools work differently, and some laws give you rights here. In the EU, AI systems that infer emotions in the workplace are prohibited, and the European Commission's guidelines treat hiring as part of the workplace for that rule. The EU's GDPR also gives you the right not to be subject to a decision based solely on automated processing that significantly affects you, with limited exceptions. In New York City, employers that use an automated tool to assess candidates must tell candidates who live in the city in advance, including which qualifications and characteristics the tool assesses. ### How should you prepare? Prepare as you would for any structured interview: know your examples, practice out loud, and set up somewhere quiet. 1. **Read the job post twice** and list the three or four things it asks for most. 2. **Pick an example for each.** Know the situation, what you did, and what happened as a result. 3. **Say your examples out loud** and time them. Some screens, including Emoti's, show a time guide for each question. 4. **Check the practical things:** a quiet room, a charged device, headphones with a microphone, and a stable connection. 5. **Choose your language.** Some screens offer several languages. Pick the one you answer best in. 6. **Read the first page carefully.** It should say how long the screen takes, whether the camera is used, and who reviews your answers. Expect questions in a few familiar areas, each with a simple shape for a good answer: | Area | Example question | A good answer includes | | --- | --- | --- | | Your work | "Tell me about a piece of work you're proud of." | One example: the problem, your part in it, the result | | Motivation | "What would make this role a clear yes for you?" | Reasons tied to this job, not to any job | | Must-haves | "Tell me about a time you were on call for a service." | A specific case, what you did, and what you learned | | Logistics | "When could you start, and what pay are you looking for?" | Your notice period, and a pay range | During the screen, take a breath before each answer. In Emoti's screens you can ask to hear a question again and pause between questions. If you are interrupted or cut off, keep going and tell the recruiter afterward. ### How do you ask for an adjustment or a human alternative? Write to the recruiter before you start, say what you need, and ask for the answer in writing. You decide how much to share about why. > Hello, I've been invited to the AI voice screen for the data engineer role. I have a stammer, and timed spoken answers put me at a disadvantage. Could I do this first round as a call with a person, or answer the questions in writing instead? Thank you. Common requests include extra time, the questions in writing as well as aloud, answering in writing, a live interview with a person, and a break partway through. Asking early gives the employer time to arrange it. The law may be on your side. In the US, the Americans with Disabilities Act requires covered employers to make reasonable accommodations for qualified applicants with disabilities, unless that would cause undue hardship. In Great Britain, the Equality Act 2010 places a duty on employers to make reasonable adjustments. In New York City, an employer's notice about an automated tool must include instructions for requesting an accommodation. In Illinois, an employer that uses AI to analyze a recorded video interview must explain how it works and get your consent first. None of this is legal advice. ### What happens to the recording? That depends on the employer and on the law where you are. Before you start, ask who can listen, how long the recording is kept, and how to have it deleted. Three questions cover it: "Who will listen to or read my answers?", "How long will you keep the recording?" and "How do I ask for it to be deleted?" In Illinois, if an employer used AI to analyze your recorded video interview, it must delete the video within 30 days of your request. In the EU, you can ask any organization for a copy of the personal data it holds about you (GDPR Article 15). In Emoti's screens, the first page tells you how long the screen takes, whether the camera is required, optional or off, the language and voice you will hear, and that people at the company review every answer before any decision. It doesn't state how long recordings are kept, so ask the employer. ### What if the screen goes wrong? Tell the recruiter in writing the same day: what happened, at which question, and what you would like to do about it. Dropped connections, a misheard question or a noisy background happen. A short, factual note such as "My connection dropped during question three, and my answer was cut off" gives the recruiter a reason to read your answers with that in mind, or to offer you another attempt. If you think the screen misheard you, say that too. Speech recognition is less accurate for some speakers than for others: one study of five commercial systems found an average word error rate of 0.35 for Black speakers and 0.19 for white speakers (Koenecke et al., 2020). A person reviewing your answers can only allow for an error they know about. Employers who use these tools well want to hear about problems, because a screen that fails a strong candidate costs them a hire. If you want to understand how one screen is designed, including where its gaps are, Emoti publishes ours in [Designing first-round voice screens candidates can trust](https://emoti.ai/research/trustworthy-voice-screens). ### Key terms - **AI voice screen**: A first-round interview in which an AI interviewer asks a fixed set of questions by voice and each answer is rated against a rubric the hiring team wrote. In Emoti, a recruiter reviews the results and makes every decision. - **Rubric**: A written guide for rating answers: for each question or competency, what a strong, middling and weak answer contains. - **Reasonable adjustment**: A change to a hiring process or job that removes a disadvantage for a disabled person, such as extra time, a different format or a human interviewer. US law calls it a reasonable accommodation. - **Automated employment decision tool**: A term from New York City's Local Law 144 for a computer-based tool that uses machine learning, statistical modeling, data analytics or artificial intelligence to substantially assist or replace discretionary hiring or promotion decisions. ### Questions #### Will an AI voice screen judge my accent? It shouldn't score your accent, but speech recognition can mishear it, which can change how an answer is read. If you are worried, ask for a human or written alternative, and tell the recruiter if you think you were misheard. #### Can I use notes during an AI voice screen? Ask the employer. If notes are allowed, keep them to a few words per example so that you are talking, not reading. #### Do I have to turn my camera on? Only if the employer requires it. Emoti's screens can be audio only, camera optional or camera required, and the first page tells you which. #### Can I take the screen in another language? If the employer offers more than one language, you can choose on the first page. Emoti's screens can run in up to ten languages. #### Can I refuse an AI screen? You can ask for an alternative, and in some places you have a right to an adjustment. In Illinois, an employer may not use AI to evaluate a recorded video interview without your consent. ### Sources - European Parliament and Council of the European Union (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2024/1689/oj - European Commission (2025). Commission publishes the Guidelines on prohibited artificial intelligence (AI) practices, as defined by the AI Act. European Commission. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act - Dushi, D. (2026). Red Lines under EU AI Act: Unpacking the prohibition of emotion recognition in the workplace and education institutions. Future of Privacy Forum. https://fpf.org/?p=248772 - European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Articles 15 and 22. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj - NYC Department of Consumer and Worker Protection (2023). Automated Employment Decision Tools: Frequently Asked Questions. City of New York. https://www.nyc.gov/assets/dca/downloads/pdf/about/DCWP-AEDT-FAQ.pdf - United States Congress (1990). Americans with Disabilities Act, 42 U.S.C. § 12112. United States Code, via govinfo.gov. https://www.govinfo.gov/content/pkg/USCODE-2010-title42/pdf/USCODE-2010-title42-chap126-subchapI-sec12112.pdf - UK Parliament (2010). Equality Act 2010, section 20: Duty to make adjustments. legislation.gov.uk. https://www.legislation.gov.uk/ukpga/2010/15/section/20 - Illinois General Assembly (2019). Artificial Intelligence Video Interview Act (820 ILCS 42). Illinois Compiled Statutes. https://www.ilga.gov/legislation/ilcs/ilcs3.asp?ActID=4015&ChapterID=68 - Koenecke, A., Nam, A., Lake, E., Nudell, J., Quartey, M., Mengesha, Z., Toups, C., Rickford, J. R., Jurafsky, D., & Goel, S. (2020). Racial disparities in automated speech recognition. Proceedings of the National Academy of Sciences, 117(14), 7684–7689. https://doi.org/10.1073/pnas.1915768117 ## Article: The hiring funnel numbers worth tracking > The hiring funnel numbers worth tracking are time to first contact, reply rate, screen pass-through, interview-to-offer, offer acceptance, time to fill, time to hire and source of hire, each with a written definition. Read them together, one role at a time, and be wary of small numbers, which swing widely. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For talent leaders. Canonical: https://emoti.ai/articles/hiring-funnel-metrics ### Which numbers are worth tracking? Eight numbers cover most of the funnel: time to first contact, reply rate, screen pass-through, interview-to-offer, offer acceptance rate, time to fill, time to hire and source of hire. Each needs a written definition before it is worth comparing. | Metric | Formula | What it tells you | | --- | --- | --- | | Time to first contact | Date of the first message to a candidate minus the date they entered the pipeline | How quickly you act on the people you find or who apply | | Reply rate | People who replied divided by people contacted | Whether outreach reaches the right people with the right message | | Screen pass-through | People advanced after the first screen divided by people screened | Whether sourcing and screening agree about who fits | | Interview-to-offer | Offers made divided by people who reached the final interview | Whether the final stage confirms what earlier stages found | | Offer acceptance rate | Offers accepted divided by offers made | Whether the offer, the pay and the process win people over | | Time to fill | Date an offer is accepted minus the date the role was opened | How long the team waits for the hire | | Time to hire | Date an offer is accepted minus the date the eventual hire entered the pipeline | How long the successful candidate's journey took | | Source of hire | Hires from each source divided by all hires | Where hires actually come from | Use medians for the time measures. One role that stays open for months will pull an average far away from what is typical. Write the edge cases into the definitions. Does a bounced email count as contacted? Do offers still waiting for an answer count in the acceptance rate? Is "entered the pipeline" the day someone applied, was found in sourcing, or was first contacted? Any choice is fine, as long as it doesn't change halfway through the year. In Emoti's outreach pages, for example, reply rate is replies divided by everyone whose first message has gone out, including bounces; people still waiting for their first message are left out. ### How do time to fill and time to hire differ? Time to fill measures the role, from opening to an accepted offer. Time to hire measures the person, from the day the eventual hire entered the pipeline to the day they accepted. The gap between them shows where the time went. A long time to fill with a short time to hire means the right person arrived late: the problem is sourcing, or the role was opened before anyone was ready to search. If both are long, the process itself is slow. Definitions vary between organizations, so say which one you use whenever you share either number. An example with invented dates: a role opens on March 3. The eventual hire is found in sourcing on April 14 and accepts an offer on May 19. Time to fill is 77 days and time to hire is 35 days. The six weeks before the right person entered the pipeline are a sourcing question, not an interviewing one. ### How should you read the numbers together? As a chain, in order, one role at a time. Each stage's output is the next stage's input, so improving one number often just moves a problem along. | Pattern | Likely reading | What to check | | --- | --- | --- | | Reply rate rises, screen pass-through falls | Outreach is reaching more people who don't fit | The brief and the search criteria | | Pass-through high, interview-to-offer low | The screen and the final interview test different things | Whether both use the same must-haves | | Offers made, acceptance low | Pay, speed or the experience lost people at the end | The pay range, and time from final interview to offer | | Time to first contact long | A backlog between finding people and contacting them | Recruiter capacity and approval steps | | One source produces most hires | Either a strong source or a narrow one | Pass-through by source, as well as volume | Read stage counts next to rates. A 50 percent pass-through rate means something different at 4 of 8 than at 40 of 80. ### What are the common pitfalls? - **Small numbers.** With five candidates at a stage, one person moves the rate by 20 points. Wait for enough people, combine periods, or report counts instead of percentages. - **Mixing roles.** A funnel for senior engineers and one for sales development representatives have little in common. Average them and you learn about neither. - **Mixing channels.** Email and LinkedIn messages can draw replies at different rates. Report reply rate per channel before you combine them. - **Moving start dates.** If time to fill starts when the role is approved for one team and when the job is posted for another, the comparison means nothing. - **Survivorship.** Measuring only the people who reached a stage hides everyone who dropped out before it. Track withdrawals and their reasons. - **Targets that bend behavior.** Push pass-through up and screens get easier; push time to fill down and offers go to the first acceptable person. Watch the next stage whenever you push on one. - **Ignoring who passes.** A stage can look healthy overall and still filter out one group. Compare pass-through rates by group. Under the US Uniform Guidelines on Employee Selection Procedures, a selection rate for any race, sex or ethnic group below four-fifths of the rate for the group with the highest rate is generally regarded as evidence of adverse impact, though differences based on small numbers may not be (29 CFR 1607.4(D)). ### How do you set this up without a data team? Agree on the definitions, record six dates for every candidate, and review the numbers monthly, per role. 1. **Write the definitions** on one page and share it with every recruiter. 2. **Record six dates per candidate:** entered the pipeline, first contact, first screen, final interview, offer, and the offer's outcome. 3. **Record the source** for everyone who enters the pipeline, not only for hires, so you can compare pass-through by source. 4. **Review monthly per role** and quarterly across similar roles. 5. **Note every change.** When you change the brief, the outreach or the screen, write down the date, so later shifts in the numbers have an explanation. Emoti keeps a count for each role at each stage: sourced, contacted, replied, screened, interviewing and offer. Each outreach sequence shows its own reply rate, and each voice screen shows who is still to finish, who is waiting for review and what has been decided. ### What should you not do with these numbers? Don't compare yourself with someone else's benchmark until you know their definitions, their roles and their market. We don't publish benchmark figures, because a number without its definition and context invites the wrong comparison. Your own history, for the same kind of role, measured the same way, is the most useful baseline you have. Use the numbers to find questions, such as why acceptance fell for one team this quarter, rather than to grade recruiters. A recruiter on a hard role with a thin market will look slow on every chart, and the chart won't say why. ### Key terms - **Time to fill**: The time from opening a role, usually the day the requisition is approved, to the day a candidate accepts an offer. - **Time to hire**: The time from the day the eventual hire entered the pipeline, by applying or being sourced, to the day they accepted the offer. - **Pass-through rate**: The share of candidates at one stage of hiring who move to the next, for example people advanced after a screen divided by people screened. - **Offer acceptance rate**: Offers accepted divided by offers made in a period, leaving out offers still waiting for an answer. - **Source of hire**: Where each hire came from, such as sourcing, referrals, the careers page, job boards or agencies, usually reported as each source's share of hires. - **Reply rate**: Replies divided by people contacted in an outreach sequence or campaign. Teams should decide whether bounced messages count as contacted and use the same rule every time. ### Questions #### What is a good reply rate for recruiting outreach? It depends on the role, the market, the channel and how you define it, so we don't quote one. Compare a sequence with your own earlier sequences for the same kind of role, measured the same way. #### Should we measure quality of hire? Yes, carefully. Pick one or two signals you can collect consistently, such as the hiring manager's rating at six months or retention at a year, and read them per role, with small numbers in mind. #### How often should we review funnel numbers? Monthly per role is usually enough to act on, and quarterly across roles for trends. Weekly reviews mostly show noise unless volumes are large. #### Do candidate withdrawals count against pass-through? Track them separately. A candidate who withdraws didn't fail a stage, and mixing the two hides why people leave. ### Sources - US Equal Employment Opportunity Commission (1978). Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4, Information on impact. Electronic Code of Federal Regulations. https://www.ecfr.gov/current/title-29/subtitle-B/chapter-XIV/part-1607/section-1607.4 ## Article: Using AI in hiring responsibly, a checklist > Using AI in hiring responsibly means telling candidates where AI is used, keeping a person on every decision, checking outcomes for adverse impact, offering adjustments, and keeping only the data you need. Rules to know early include New York City's Local Law 144, Illinois's Artificial Intelligence Video Interview Act and the EU AI Act, which lists recruitment as a high-risk use; this checklist is not legal advice. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For talent leaders. Canonical: https://emoti.ai/articles/using-ai-in-hiring-responsibly ### What does responsible use of AI in hiring involve? Five habits, kept every time a tool touches a candidate: tell people, keep a person on every decision, check outcomes for adverse impact, offer adjustments, and collect and keep only the data you need. Use this checklist before you switch a tool on, and again every few months: 1. Candidates are told, in plain words and before the AI step, what the tool does and who reviews its output. 2. A named person makes every advance or reject decision and can see the evidence behind it. 3. Selection rates are compared by group at each stage the tool affects. 4. Candidates know how to ask for an adjustment or an alternative, and requests are answered quickly. 5. You know what data the tool collects, why, where it goes and when it is deleted. 6. You know which laws apply where you hire, and who is responsible for each duty. This article is not legal advice. Rules change, and the official pages linked below are the place to check the current position. ### How should you tell candidates? Before the AI step, say what the tool does, what it evaluates, who sees the result, and how to ask for an alternative or an adjustment. Plain words work better than legal ones: "A recorded AI interviewer will ask you five questions about your work. People on our recruiting team review every answer before any decision." Some laws set minimums. New York City's Local Law 144 requires employers and employment agencies to tell candidates who live in the city, 10 business days before using an automated employment decision tool, that the tool will be used and which job qualifications and characteristics it will assess, with instructions for requesting a reasonable accommodation. Illinois's Artificial Intelligence Video Interview Act requires employers that use AI to analyze recorded video interviews for Illinois-based positions to notify applicants, explain how the AI works and what general types of characteristics it uses, and obtain consent before the interview. ### How do you keep a person on every decision? Make sure a named person decides each advance or reject, sees the evidence the tool used, and has the time and authority to disagree with it. A tool should never reject anyone on its own. Human review only counts if it is real. A recruiter who approves 200 recommendations in ten minutes is not reviewing them. Give reviewers the underlying evidence as well as any score; track how often they disagree with the tool; and treat a disagreement rate near zero as a warning rather than a success. In the EU, the GDPR gives people the right not to be subject to a decision based solely on automated processing that produces legal effects for them or similarly significantly affects them. Where such a decision is allowed because it is necessary for a contract or the person has given explicit consent, they must still be able to obtain human intervention, give their point of view and contest the decision (Article 22). ### How do you check for adverse impact? Compare the selection rate of each group at each stage. Under the US Uniform Guidelines on Employee Selection Procedures, a selection rate for any race, sex or ethnic group that is less than four-fifths (80 percent) of the rate for the group with the highest rate is generally regarded as evidence of adverse impact (29 CFR 1607.4(D)). An illustrative calculation, with made-up numbers: | Group | Applied | Advanced | Selection rate | Ratio to highest rate | | --- | --- | --- | --- | --- | | Group A | 100 | 30 | 30% | 1.00 | | Group B | 90 | 18 | 20% | 0.67 | Group B's ratio of 0.67 is below 0.80, so the stage would be flagged for a closer look. The guideline is a rule of thumb, not a verdict. The same section says smaller differences can still be adverse impact when they are significant in both statistical and practical terms, and larger ones may not be when they rest on small numbers. New York City's law turns this into a duty for covered tools: an independent bias audit within one year before use, with selection or scoring rates and impact ratios by sex, by race and ethnicity, and by intersectional categories, and a published summary of the results. ### How should you handle adjustments and data? Tell candidates how to ask for an adjustment and answer quickly, and collect only the data a decision needs, for only as long as you need it. #### Adjustments An AI screening step can quietly assume a candidate who hears well, speaks fluently and is comfortable being timed. That is not everyone. In the US, the Americans with Disabilities Act counts not making reasonable accommodations to the known limitations of a qualified applicant with a disability as discrimination, unless the accommodation would impose an undue hardship. In Great Britain, the Equality Act 2010 places a duty on employers to make reasonable adjustments. Offer a real alternative, such as a live interview, written answers or extra time, and put a sentence in every invitation: "If you'd like an adjustment or a different format, reply to this email." #### Data For each tool, write down what it collects (audio, video, transcripts, scores), why, who can see it, and when it is deleted. The GDPR requires personal data to be limited to what is necessary for its purpose and kept no longer than necessary (Article 5). Illinois's video interview law limits sharing applicant videos to people whose expertise or technology is needed to evaluate the applicant, and requires employers to delete an applicant's interviews within 30 days of the applicant's request. ### Which rules do teams meet first? Three are a good place to start: New York City's Local Law 144, the Illinois Artificial Intelligence Video Interview Act and the EU AI Act. Other places have their own rules, so check where you hire. | Rule | Where it applies | What it requires | Status | | --- | --- | --- | --- | | Local Law 144 of 2021 | Automated employment decision tools used to assess candidates for jobs in New York City | An independent bias audit within a year before use, a published summary, and notice to candidates 10 business days before use | Enforced since July 5, 2023 | | Artificial Intelligence Video Interview Act (820 ILCS 42) | AI analysis of recorded video interviews for positions based in Illinois | Notice, an explanation of how the AI works, consent, limits on sharing and deletion on request; demographic reporting for employers that rely solely on AI to choose who gets an in-person interview | In effect since January 1, 2020; reporting since January 1, 2022 | | EU AI Act (Regulation (EU) 2024/1689) | AI systems used in the EU, including for recruitment and selection | Recruitment and selection systems are listed as high-risk in Annex III; AI that infers emotions in the workplace is prohibited | High-risk obligations for Annex III systems apply from December 2, 2027, under amending Regulation (EU) 2026/1744; the prohibition has applied since February 2, 2025 | A few details matter. New York City's Department of Consumer and Worker Protection says the law's requirements apply when a tool is used to assess people who have applied for a specific position, not when it is used to scan a resume bank, reach out to potential candidates or invite applications. The vendor of a tool is not responsible for its bias audit under that law; the employer is. And the European Commission's guidelines on prohibited practices treat hiring as part of the workplace for the emotion recognition ban. #### Where Emoti fits Emoti is designed around the same habits, and it doesn't do the employer's legal work. Every ranking shows its evidence, criterion by criterion. Emoti never advances or rejects anyone; a person makes each decision. Voice screens tell candidates that an AI interviewer asks the questions and that people review every answer, and they rate the content of answers against the team's rubric, not the candidate's face or tone of voice. Our research papers describe each method's limits, including gaps we haven't closed yet, such as the lack of a separate consent step in voice screens: see [Evidence-first candidate matching](https://emoti.ai/research/evidence-first-matching) and [Designing first-round voice screens candidates can trust](https://emoti.ai/research/trustworthy-voice-screens). Notices, audits and decisions about adjustments remain the employer's responsibility. ### Key terms - **Adverse impact**: A substantially lower selection rate for one group, defined by race, sex or ethnicity, compared with another at a step in the hiring process. - **Four-fifths rule**: A guideline in the US Uniform Guidelines on Employee Selection Procedures: a selection rate for any race, sex or ethnic group below four-fifths of the highest group's rate is generally regarded as evidence of adverse impact. - **Automated employment decision tool**: A term from New York City's Local Law 144 for a computer-based tool that uses machine learning, statistical modeling, data analytics or artificial intelligence to substantially assist or replace discretionary hiring or promotion decisions. - **Bias audit**: An impartial evaluation of a hiring tool's outcomes by group. Under New York City's Local Law 144, an independent auditor calculates selection or scoring rates and impact ratios by sex, race and ethnicity, and their intersections. - **Reasonable adjustment**: A change to a hiring process or job that removes a disadvantage for a disabled person, such as extra time, a different format or a human interviewer. US law calls it a reasonable accommodation. ### Questions #### Does keeping a person in the loop remove legal risk? No. Anti-discrimination law applies to decisions people make with a tool's help, and a reviewer who rubber-stamps a tool's output adds little protection. Review helps when the person sees the evidence and can disagree. #### How often should we check AI hiring tools for adverse impact? Whenever a tool or its settings change, and on a regular schedule, such as every quarter, at each stage the tool affects. With small numbers, look at longer periods rather than single months. #### Can a bias audit use inferred demographic data? Not under New York City's rules. Its FAQ says imputed or inferred demographic data cannot be used to conduct a bias audit, and that test data can be used when there isn't enough historical data. #### What should we ask an AI hiring vendor? What the tool evaluates and how, what data it uses and keeps, how a person reviews its output, what bias testing has been done and on what data, and how candidates can ask for an alternative. ### Sources - US Equal Employment Opportunity Commission (1978). Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4, Information on impact. Electronic Code of Federal Regulations. https://www.ecfr.gov/current/title-29/subtitle-B/chapter-XIV/part-1607/section-1607.4 - NYC Department of Consumer and Worker Protection (2023). Automated Employment Decision Tools (AEDT). City of New York. https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page - NYC Department of Consumer and Worker Protection (2023). Automated Employment Decision Tools: Frequently Asked Questions. City of New York. https://www.nyc.gov/assets/dca/downloads/pdf/about/DCWP-AEDT-FAQ.pdf - Illinois General Assembly (2019). Artificial Intelligence Video Interview Act (820 ILCS 42). Illinois Compiled Statutes. https://www.ilga.gov/legislation/ilcs/ilcs3.asp?ActID=4015&ChapterID=68 - European Parliament and Council of the European Union (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2024/1689/oj - European Parliament and Council of the European Union (2026). Regulation (EU) 2026/1744 amending Regulations (EU) 2024/1689, (EU) 2018/1139 and (EU) 2023/1230 (Digital Omnibus on AI). Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2026/1744/oj - European Commission (2025). Commission publishes the Guidelines on prohibited artificial intelligence (AI) practices, as defined by the AI Act. European Commission. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act - Dushi, D. (2026). Red Lines under EU AI Act: Unpacking the prohibition of emotion recognition in the workplace and education institutions. Future of Privacy Forum. https://fpf.org/?p=248772 - European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Articles 5 and 22. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj - United States Congress (1990). Americans with Disabilities Act, 42 U.S.C. § 12112. United States Code, via govinfo.gov. https://www.govinfo.gov/content/pkg/USCODE-2010-title42/pdf/USCODE-2010-title42-chap126-subchapI-sec12112.pdf - UK Parliament (2010). Equality Act 2010, section 20: Duty to make adjustments. legislation.gov.uk. https://www.legislation.gov.uk/ukpga/2010/15/section/20 ## Article: What to ask in a first-round screen, and what to leave out > A first-round screen should confirm motivation, check each must-have with one evidence question, and settle location, notice period and pay expectations, all in about 25 minutes. Leave out questions about protected characteristics, and where salary history bans apply, as in California, New York State and New York City, ask what the candidate expects to earn rather than what they earn now. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For recruiters. Canonical: https://emoti.ai/articles/first-round-screen-questions ### What is a first-round screen for? A first-round screen decides whether both sides want a full interview: does the person want this job, do they meet the must-haves, and do the practical details work? It is not the place to test depth. That belongs in later interviews, where people who know the work can probe it. A first screen that tries to do everything runs long, and the candidate leaves knowing less about the job than when they arrived. Treat it as a short structured interview. Ask every candidate the same questions in the same order, and decide in advance what a good answer contains. That makes candidates comparable, and it is fairer to the people you speak to late in the week, when your memory of the first calls has faded. ### How long should each part take? About 25 minutes covers it, if you plan the time: | Part | Minutes | What you come away with | | --- | --- | --- | | Open and set the agenda | 2 | A candidate who knows what is coming | | Motivation | 4 | Why they are open to moving, in their own words | | Must-have checks | 9 | One evidence answer for each must-have | | Logistics | 5 | Location, notice period, pay expectations, other processes | | Their questions | 3 | What matters to them | | Close | 2 | Agreed next steps and a date | If you run short of time, cut the motivation section, not the must-haves or the close. ### Which questions confirm motivation? Ask why they are open to a move now and what would make this role a clear yes, then listen for specifics. - "What made you open to a conversation now?" - "What would make this role a clear yes for you?" - "What would make you stay where you are?" Specific answers tied to the work ("I want to own a product end to end") tell you more than general ones ("I'm looking for growth"). The last question is easy to skip and worth asking: it tells you what you would be competing with at offer time, including a counteroffer. ### How do you check must-haves in a short call? Ask one evidence question per must-have, phrased to draw out a specific example, and compare the answer with the evidence you agreed counts. | Must-have | Question | Evidence that counts | | --- | --- | --- | | Has owned a production service | "Tell me about a service you were on call for. What broke last, and what did you do?" | A named service, a real incident, their own actions | | Has managed people | "How many people report to you now, and what was the last hard conversation you had with one of them?" | A number and a specific situation | | Enterprise sales | "Walk me through the largest deal you closed in the last two years." | Deal size, length, who they sold to, their part in it | Don't grade delivery. Nerves, an accent and a quiet voice are not evidence about the must-haves. If an answer is thin, use the same neutral follow-up you would use with anyone: "Can you give me a specific example?" ### Which logistics belong in the first call? Location and work pattern, the right to work where the job is, the notice period and earliest start date, pay expectations, and other processes and their timing. - **Location:** state the requirement and ask whether it works. "The role is in Berlin with two office days a week. Does that work for you?" - **Right to work:** if you can't sponsor a visa, say so and ask whether they would need sponsorship. Don't ask where they are from. - **Notice period:** "How much notice do you need to give, and when could you start?" - **Pay:** give the range first, then ask about expectations. "The base range is €95,000 to €115,000. How does that compare with what you're looking for?" - **Other processes:** "Are you talking to other companies, and is there a date we should know about?" Giving the pay range first saves both sides a second call when the numbers don't meet. In California, an employer must give an applicant the pay scale for the position on reasonable request. ### What should you leave out? Anything about protected characteristics or obvious stand-ins for them, and in many places, the candidate's pay history. In the US, the Equal Employment Opportunity Commission advises that pre-employment questions be limited to what is essential for judging whether someone is qualified. It says questions that would reveal race, sex, national origin, age or religion should generally be avoided, because they may be used as evidence of intent to discriminate. Under the Americans with Disabilities Act, employers may not ask applicants whether they have a disability, or about its nature or severity, before an offer, though they may ask about the ability to perform job-related functions. In Great Britain, the Equality Act 2010 restricts questions about an applicant's health before a job offer, with some exceptions. So leave out questions such as: - "Where are you from?" or "Is that accent from…?" - "Do you have children?" or "Are you planning a family?" - "When did you graduate?" asked as a way to estimate age. - "Do you have any health conditions?" - "What does your partner do?" Salary history is the other big one. Several US states and cities ban asking about it. Three examples: - **California** (Labor Code section 432.3): employers may not seek an applicant's salary history, personally or through an agent, or rely on it in deciding whether to make an offer or what salary to offer. - **New York State** (Labor Law section 194-a, in effect since January 6, 2020): employers may not ask applicants about their current or past pay, or rely on it. - **New York City** (in effect since October 31, 2017): employers may not ask about an applicant's salary history in job ads, on applications or in interviews. Ask about expectations instead. Official guidance in both California and New York State says employers may ask what pay an applicant expects. The rules differ in other places, and this article is not legal advice, so check the rules where you and the candidate are. ### How do you close the call? Play back what you heard, say exactly what happens next and when, and send a short written note the same day. 1. **Play back the logistics.** "So you'd need six weeks' notice, and the range works for you." 2. **Name the next step and a date.** "I'll share my notes with the hiring manager today, and you'll hear from me by Thursday either way." 3. **Ask what they need.** "Is there anything that would help you decide?" 4. **Write it down.** A short email with the next step and the date gives the candidate something to hold you to. Then write your notes before the next call. Emoti's note template for recruiter screens has a similar shape: why they're looking, current role, fit with the brief, compensation and logistics, and next steps. The same structure can also run as an AI voice screen that candidates take when it suits them and a recruiter reviews; how that works, and its limits, is in [Designing first-round voice screens candidates can trust](https://emoti.ai/research/trustworthy-voice-screens). ### Key terms - **Phone screen**: A short first conversation that checks a candidate's motivation, must-haves and logistics before a full interview. It can be run by a recruiter or as an AI voice screen. - **Notice period**: The time a person must work, or be paid for, after resigning from their current job before they can start a new one, set by their contract or local law. - **Counteroffer**: An offer a candidate's current employer makes to keep them after they receive or accept another offer, usually more pay or a new role. - **Must-have criterion**: A requirement a team would turn down an otherwise excellent candidate for lacking. In Emoti, must-haves count double, and missing one caps a candidate's score at 54. - **Structured interview**: An interview in which every candidate is asked the same job-related questions and every answer is rated against the same written standard. ### Questions #### What if a candidate volunteers their current salary? In California and New York State, an employer may consider salary history that an applicant discloses voluntarily and without prompting. Rules elsewhere differ. A pay range set before the call makes the number less relevant anyway. #### Is it OK to record a screening call? Tell the candidate before you record and ask whether that is all right; in some places you need everyone's consent. Interview notes that hold up in a debrief covers recording and consent in more detail. #### Who should run the first-round screen? Usually the recruiter, or an AI voice screen that a recruiter reviews. Either way, the hiring manager should agree the must-haves and the evidence for each one beforehand. #### How soon should candidates hear back after a screen? Give a date on the call and keep it, even when the answer is no. A quick, clear no is better for the candidate than silence. ### Sources - US Equal Employment Opportunity Commission Prohibited Employment Policies/Practices. US Equal Employment Opportunity Commission. https://www.eeoc.gov/prohibited-employment-policiespractices - United States Congress (1990). Americans with Disabilities Act, 42 U.S.C. § 12112. United States Code, via govinfo.gov. https://www.govinfo.gov/content/pkg/USCODE-2010-title42/pdf/USCODE-2010-title42-chap126-subchapI-sec12112.pdf - UK Parliament (2010). Equality Act 2010, section 60: Enquiries about disability and health. legislation.gov.uk. https://www.legislation.gov.uk/ukpga/2010/15/section/60 - Labor Commissioner's Office California Equal Pay Act and Pay Transparency Law (Labor Code sections 432.3 and 1197.5): Frequently Asked Questions. California Department of Industrial Relations. https://www.dir.ca.gov/dlse/California_Equal_Pay_Act.htm - New York State (2020). Salary History Ban - What You Need To Know. ny.gov. https://www.ny.gov/salary-history-ban/salary-history-ban-what-you-need-know - NYC Commission on Human Rights (2017). NYC Law: Salary History Question Ban. City of New York. https://www.nyc.gov/site/cchr/media/salary-history.page ## Article: Writing outreach candidates reply to > Good recruiting outreach opens with one specific line about the person's own work, states the role, team, location and pay range, and asks for one small next step. Follow-ups should each add something new, come a few days apart and stop at the first reply, and every message should make it easy to opt out. Published by Emoti Research on 2026-11-01, updated 2026-11-01. Article. For recruiters. Canonical: https://emoti.ai/articles/outreach-candidates-reply-to ### What makes a first line personal? A personal first line names something specific the person did in their own work and connects it to the role, in one sentence. Compare these two openings: > I came across your impressive profile and think you'd be a great fit for an exciting opportunity. > You rebuilt the event pipeline at your company so that late carrier data stopped breaking forecasts, and that is exactly the problem our platform team is hiring for. The first could be sent to anyone, so it reads as if it was. The second could only be sent to one person, and it explains why you are writing. Good material for that sentence comes from the person's own descriptions of their roles, talks they have given, things they have written, and open-source work. Skip compliments that say nothing ("impressive background"), and facts about them rather than their work. A few openings in the same spirit, for different roles: - **A product designer:** "Your case study on simplifying the dispatch screen showed exactly the trade-offs our planning team argues about every week." - **An account executive:** "You built the first enterprise pipeline at your last company from nothing, which is the job here too." - **A data engineer:** "Your talk on backfilling late events without double counting describes the problem our analytics team is stuck on." ### Where is the line between personal and intrusive? Use work facts the person published for professional reasons, and nothing else. Their job, their projects and their public talks are fair. Their family, their location history, their other social accounts and anything that suggests you have been watching them are not. A useful test: if they asked "how did you know that?", would you be comfortable showing them where you found it? If the honest answer is "I combined three sources", leave it out. Be careful with sensitive signals too. A profile that shows someone is open to work can tell you whom to contact, but it doesn't belong in the message. "I saw you're looking" tells the reader you noticed something they may not want discussed. ### What facts does a candidate need to decide? The role and its level, the team and what it works on, the location and work pattern, the pay range, and who you are. | Fact | Why it matters | Example | | --- | --- | --- | | Role and level | So they can tell whether it is a step up | Senior backend engineer, individual contributor | | Team and work | So they can picture the job | The payments team, which moves money between shippers and carriers | | Location | So they can rule it in or out | Berlin with two office days a week, or remote within Europe | | Pay range | So nobody spends a call on numbers that don't meet | €95,000 to €115,000 base | | Who you are | So they know who will answer | Your name and role, and who the hiring manager is | Leaving out the pay range asks the candidate to spend a call finding out whether the job is even possible for them. ### How should you make the ask? Ask for one small, specific next step that is easy to say yes to. "Would you be open to a 20-minute call next week?" is easy to answer. "Let me know your thoughts" isn't. Offer an alternative for people who would rather read first: "If you'd prefer to read before talking, I can send the brief." Don't stack asks. A call, a request for referrals and a link to an application form in one message dilute each other. Keep the whole message short enough to read on a phone without scrolling: the first line, the facts, the ask and your name. Put together, a first message can be this short (the names and numbers are invented): > Hi Sam, your talk on backfilling late events without double counting describes the problem our analytics team is stuck on. I'm Alex, and I recruit for Example Co.'s data platform team: a senior data engineer role, individual contributor, reporting to our head of data. It's in Berlin with two office days a week, or remote within Europe, and the base range is €90,000 to €105,000. Would you be open to a 20-minute call next week? If you'd rather read first, I can send the brief, and if you'd rather not hear from me again, reply and say so. ### How far apart should follow-ups be? A few days apart at first, then longer, with each one adding something new, and no more than three in total. | Day | Step | What it adds | | --- | --- | --- | | 1 | Email | The first line, the facts and one ask | | 3 | LinkedIn message | A short note, in case email isn't where they read | | 6 | Email | One new detail: the team, the process or the hardest problem | | 10 | Email | A clear last note that says it is the last | A follow-up that only says "just bumping this" asks for attention without giving a reason. Each one should earn its place with a fact the first message didn't have room for. Stop as soon as someone replies, whatever the reply says, and don't send the same message on two channels on the same day. ### How do you make opting out easy? Tell people plainly how to stop hearing from you, honor it everywhere, and keep a record. One sentence is enough: "If you'd rather not hear from me again, reply and say so, and I won't write again." When someone says no, record it where everyone on your team can see it, so that a colleague doesn't contact them next month about a different role. A person who said "not now" may welcome a note in six months. A person who said "please stop" should not get one. There is a legal side in some places. In the EU, when you collect someone's personal data from somewhere other than the person, the GDPR requires you to tell them, among other things, where the data came from and how you will use it, and if you use it to contact them, to do so at the latest when you first contact them (Article 14). This article is not legal advice; check the rules where you recruit. ### How does Emoti draft first lines for approval? Emoti drafts one first line per person from facts in their profile, a recruiter approves each line, and nothing is sent until the line is approved and the sequence is active. In version 1.0, first lines are written by fixed rules rather than a language model, so every clause can be traced to the profile: the description of their current role, skills that overlap with the role's brief, background in a related industry, or a professional signal such as speaking at conferences or maintaining open-source projects. Emoti never puts "open to work" into a message, and it marks the lines it wrote. Emails go out on weekdays within a sending window in each recipient's time zone (8:30 a.m. to 5:30 p.m. by default), under a daily limit per mailbox (60 by default). Any reply stops the sequence for that person. Emoti doesn't yet add an unsubscribe link or keep a workspace-wide do-not-contact list, so the opt-out sentence above matters. The details, including the gaps, are in [Personal outreach with a person in the loop](https://emoti.ai/research/outreach-with-approval). ### Key terms - **First line**: In outreach, the opening sentence of the first message, written for one person from their own work. Emoti drafts one per person and a recruiter approves it before anything is sent. - **Outreach sequence**: A planned series of messages and tasks, such as emails, LinkedIn messages, calls and waits, sent to a candidate over days or weeks until they reply or the sequence ends. - **Reply rate**: Replies divided by people contacted in an outreach sequence or campaign. Teams should decide whether bounced messages count as contacted and use the same rule every time. - **Passive candidate**: A person who isn't actively looking for a new job but might consider the right one. Sourcing and outreach are how teams reach them. ### Questions #### Should I mention pay in the first outreach message? Yes, as a range. It saves both sides a call when the numbers don't meet, and in some places job postings must include a pay range anyway. #### Is it OK to say how I found someone? Yes, and it helps. Naming the talk, project or profile you read shows the first line is real. In the EU, telling people where their data came from is also required when you didn't collect it from them. #### Should I use email or LinkedIn for outreach? Start on the channel the person seems most active on, and use the other for one follow-up. Don't send the same message on both on the same day. #### What should I do when someone replies "not now"? Thank them, ask when would be a better time, and record the date. In Emoti their reply stops the sequence, so set your own reminder to check back. ### Sources - European Parliament and Council of the European Union (2016). Regulation (EU) 2016/679 (General Data Protection Regulation), Article 14. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj - Labor Commissioner's Office California Equal Pay Act and Pay Transparency Law (Labor Code sections 432.3 and 1197.5): Frequently Asked Questions. California Department of Industrial Relations. https://www.dir.ca.gov/dlse/California_Equal_Pay_Act.htm ## Glossary of recruiting terms > Plain definitions of recruiting and hiring terms, from Emoti. Canonical: https://emoti.ai/glossary ### Adverse impact A substantially lower selection rate for one group, defined by race, sex or ethnicity, compared with another at a step in the hiring process. In the US it is often first checked with the four-fifths rule. Adverse impact can happen without any intent to discriminate, which is why outcomes have to be measured rather than assumed. Check each stage of the process separately, and be careful when the numbers are small. See also: https://emoti.ai/articles/using-ai-in-hiring-responsibly ### AI voice screen A first-round interview in which an AI interviewer asks a fixed set of questions by voice and each answer is rated against a rubric the hiring team wrote. In Emoti, a recruiter reviews the results and makes every decision. Candidates usually take it from a link at a time that suits them. A well-designed screen tells candidates an AI is asking the questions, explains who reviews the answers, and offers a way to ask for an adjustment. See also: https://emoti.ai/research/trustworthy-voice-screens ### Anchored rating scale A rating scale in which each point is described by an example or description of the answer that earns it, so interviewers rate answers against the same standard. Reviews of the research find that anchored scales make structured interview ratings more reliable and more valid. Anchoring every point on the scale, not only the ends, gives raters the most guidance. See also: https://emoti.ai/articles/structured-interviews ### Applicant tracking system Software that stores job openings, applications and each candidate's progress through the hiring stages, often called an ATS. It is usually the system of record for hiring decisions. ### Automated employment decision tool A term from New York City's Local Law 144 for a computer-based tool that uses machine learning, statistical modeling, data analytics or artificial intelligence to substantially assist or replace discretionary hiring or promotion decisions. Employers using one to assess candidates for jobs in New York City must have an independent bias audit done within a year before use, publish a summary of the results, and notify candidates who live in the city 10 business days before use. See also: https://emoti.ai/articles/using-ai-in-hiring-responsibly ### Bias audit An impartial evaluation of a hiring tool's outcomes by group. Under New York City's Local Law 144, an independent auditor calculates selection or scoring rates and impact ratios by sex, race and ethnicity, and their intersections. See also: https://emoti.ai/articles/using-ai-in-hiring-responsibly ### Boolean search A search that combines keywords with operators such as AND, OR and NOT, and quotation marks for exact phrases, to find profiles or resumes. It is precise but literal: it finds the words people used, not necessarily the skills they have. ### Calibration Aligning interviewers, or a recruiter and a hiring manager, on what a good candidate looks like, usually by rating the same examples together and discussing the differences. In sourcing, calibration often means reviewing the first few profiles together before searching further. ### Candidate experience How candidates perceive an employer's hiring process, from the first message to the final decision, including its speed, clarity, fairness and respect. ### Competency A skill, behavior or kind of knowledge the job requires, defined clearly enough to be assessed, such as owning production systems or negotiating contracts. See also: https://emoti.ai/articles/structured-interviews ### Counteroffer An offer a candidate's current employer makes to keep them after they receive or accept another offer, usually more pay or a new role. See also: https://emoti.ai/articles/first-round-screen-questions ### Debrief The meeting after interviews where interviewers share their ratings and evidence and the hiring team decides. It works best when every interviewer has recorded ratings independently beforehand. See also: https://emoti.ai/articles/interview-notes-that-hold-up ### Evidence In Emoti, the line of profile or transcript text behind each judgment, shown next to the judgment so a person can check it. In sourcing, every criterion on the fit strip carries an evidence line, such as the employer where a skill was used or a note that the skill isn't in the profile. In interview notes, evidence is a quote linked to its moment in the call. See also: https://emoti.ai/research/evidence-first-matching ### First line In outreach, the opening sentence of the first message, written for one person from their own work. Emoti drafts one per person and a recruiter approves it before anything is sent. See also: https://emoti.ai/research/outreach-with-approval ### Fit strip Emoti's row of marks, one per search criterion, showing whether a candidate meets, partly meets, misses or lacks information for each criterion, with the evidence behind each mark. Met is worth 1, partly met 0.5, not enough information 0.25 and missing 0. Must-haves count double, a missing must-have caps the score at 54 out of 100, and a partly met must-have caps it at 74. See also: https://emoti.ai/research/evidence-first-matching ### Four-fifths rule A guideline in the US Uniform Guidelines on Employee Selection Procedures: a selection rate for any race, sex or ethnic group below four-fifths of the highest group's rate is generally regarded as evidence of adverse impact. It is a rule of thumb, not a legal verdict. The guidelines note that smaller differences can still matter when they are statistically and practically significant, and larger ones may not when they rest on small numbers. See also: https://emoti.ai/articles/using-ai-in-hiring-responsibly ### Hiring brief A short, plain-language description of who a team wants to hire and how it will recognize them: must-haves, nice-to-haves, evidence, location, level, pay range and deal-breakers. See also: https://emoti.ai/articles/writing-a-hiring-brief ### Intake meeting The meeting where a recruiter and hiring manager agree on a role before the search starts: what the person will do, the must-haves, where to look and the interview process. See also: https://emoti.ai/articles/writing-a-hiring-brief ### Job description A document describing a job's purpose, responsibilities and requirements. It is usually written for candidates and internal records, and is broader than a hiring brief. ### Job post A public advertisement for an open role, published on a careers page or job board and written to attract applicants. Some places require it to include a pay range. California, for example, requires employers with 15 or more employees to include the pay scale in job postings. ### Look-alike search In Emoti, a search that builds criteria from one person's profile, such as the same discipline, region, years of experience and top skills, and ranks others against them. The criteria are visible and editable, which matters because copying the shape of one profile can also copy its blind spots. See also: https://emoti.ai/research/evidence-first-matching ### Market map In Emoti, a view of where people in one discipline and region work, grouped by company, with how many there are, how many are senior, their median years of experience and their most common skills. ### Must-have criterion A requirement a team would turn down an otherwise excellent candidate for lacking. In Emoti, must-haves count double, and missing one caps a candidate's score at 54. Keep the list short: each extra must-have shrinks the pool. A good test is to ask whether you would really turn down a strong candidate who lacked it. See also: https://emoti.ai/articles/writing-a-hiring-brief ### Nice-to-have criterion A requirement that would make a candidate stand out but whose absence wouldn't rule them out. In Emoti, nice-to-haves count once and never cap a score. See also: https://emoti.ai/articles/writing-a-hiring-brief ### Notetaker Software that joins a call, records and transcribes it, and drafts notes. Emoti's notetaker writes notes in the template chosen for each kind of call and, by default, tells everyone the call is recorded. See also: https://emoti.ai/research/grounded-interview-notes ### Notice period The time a person must work, or be paid for, after resigning from their current job before they can start a new one, set by their contract or local law. See also: https://emoti.ai/articles/first-round-screen-questions ### Offer acceptance rate Offers accepted divided by offers made in a period, leaving out offers still waiting for an answer. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Outreach sequence A planned series of messages and tasks, such as emails, LinkedIn messages, calls and waits, sent to a candidate over days or weeks until they reply or the sequence ends. In Emoti, any reply stops the sequence for that person, and emails go out only on weekdays, within a sending window and under a daily limit per mailbox. See also: https://emoti.ai/articles/outreach-candidates-reply-to ### Panel interview An interview in which several interviewers meet a candidate together, or a set of interviews with different interviewers. Each panel member should rate the candidate independently before any discussion. See also: https://emoti.ai/articles/structured-interviews ### Pass-through rate The share of candidates at one stage of hiring who move to the next, for example people advanced after a screen divided by people screened. Comparing pass-through rates by group at each stage is one of the simplest ways to spot adverse impact. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Passive candidate A person who isn't actively looking for a new job but might consider the right one. Sourcing and outreach are how teams reach them. See also: https://emoti.ai/articles/outreach-candidates-reply-to ### Phone screen A short first conversation that checks a candidate's motivation, must-haves and logistics before a full interview. It can be run by a recruiter or as an AI voice screen. See also: https://emoti.ai/articles/first-round-screen-questions ### Quality of hire A measure of how well hires turn out, using signals such as performance ratings, hiring manager satisfaction or retention after a set period. There is no single standard definition. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Reasonable adjustment A change to a hiring process or job that removes a disadvantage for a disabled person, such as extra time, a different format or a human interviewer. US law calls it a reasonable accommodation. Great Britain's Equality Act 2010 places a duty on employers to make reasonable adjustments, and the US Americans with Disabilities Act requires reasonable accommodations for qualified applicants unless they would cause undue hardship. See also: https://emoti.ai/articles/ai-voice-screens-for-candidates ### Reply rate Replies divided by people contacted in an outreach sequence or campaign. Teams should decide whether bounced messages count as contacted and use the same rule every time. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Requisition An approved request to hire for a role, usually with a budget, level and headcount, often shortened to req. Time to fill is usually counted from the day it opens. ### Rubric A written guide for rating answers: for each question or competency, what a strong, middling and weak answer contains. See also: https://emoti.ai/articles/structured-interviews ### Scorecard The record of one interviewer's or one screen's ratings for a candidate, competency by competency or question by question, with the evidence for each rating. See also: https://emoti.ai/articles/interview-notes-that-hold-up ### Source of hire Where each hire came from, such as sourcing, referrals, the careers page, job boards or agencies, usually reported as each source's share of hires. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Sourcing Finding and identifying people who could fit a role, especially people who haven't applied, through searches, referrals and research. See also: https://emoti.ai/research/evidence-first-matching ### Structured interview An interview in which every candidate is asked the same job-related questions and every answer is rated against the same written standard. Research consistently finds that structured interviews predict job performance better than unstructured conversations, and a 2022 re-analysis ranked them first among common selection methods. See also: https://emoti.ai/articles/structured-interviews ### Talent pool A saved group of potential candidates, such as past finalists or people found in sourcing, kept for current or future roles. ### Time to fill The time from opening a role, usually the day the requisition is approved, to the day a candidate accepts an offer. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Time to hire The time from the day the eventual hire entered the pipeline, by applying or being sourced, to the day they accepted the offer. Read it next to time to fill. A long time to fill with a short time to hire means the right person arrived late, which usually points to sourcing rather than the interview process. See also: https://emoti.ai/articles/hiring-funnel-metrics ### Transcript highlight In Emoti, a marked moment in a call's transcript with a type (strength, concern, question, logistics or motivation), a short title, a quote and a time that opens the transcript at that moment. See also: https://emoti.ai/research/grounded-interview-notes