How Emoti uses AI

Emoti uses AI for the reading, drafting and summarizing that fill a recruiter's day, so people can spend their time talking to candidates and making decisions. The AI suggests. People decide.

Where AI and the agent help

• Briefing the agent. On Home, you describe a role, paste a job description or ask about a candidate. The agent replies with a short plan and up to three things it has prepared: a sourcing search, an outreach sequence draft, an AI screen draft or a page to open. It never sends, schedules or changes anything by itself.
• Turning a brief into criteria. A search reads your description into criteria such as role, location, years of experience, skills and industry, each marked required or nice to have. You can edit every one.
• Personal first lines. The agent writes an opening line for each person in an outreach sequence, from their profile. Nothing is sent until a recruiter approves it.
• AI voice screens. An AI interviewer asks the hiring team's questions, in the candidate's language and on their schedule. Each answer is rated Weak, Mixed, Good or Strong against the team's rubric and summarized for the team.
• Notes and Ask. The notetaker writes notes from a recorded call, following the template for that kind of call, with a short summary and highlights. Ask answers questions about a call and quotes the transcript with time links.
• Insights. Saved questions search the transcripts, highlights and AI notes of many calls and quote the moments that match.
• Help. Emoti Genie answers questions from these help articles and links the ones it used. See Emoti Genie, the assistant.

What follows fixed rules

Some parts of Emoti deliberately don't use an AI model, so their results are predictable and easy to explain:

• Fit scores. Every person is checked against each criterion with the same rules, and the fit strip shows the evidence behind each mark. Must-haves count double, and missing one keeps a score below 55. In Sourcing, partly meeting one keeps it below 75.
• Job post checks. The checks for inclusive language, clarity, requirements, pay transparency and length rerun as you type. The same text always gets the same result.

People decide and approve what's sent

AI output in Emoti is a starting point. The agent prepares drafts and doesn't send them: first lines wait for a recruiter's approval, a sequence sends only after someone starts it, and a screen invite goes out only when someone on the team invites the candidate. A reply stops the sequence for that person. There's no automatic rejection either: a fit score or a screen rating puts people in order for review, and a person decides who advances, who's held and who's rejected.

Candidates are told when AI is involved. An AI screen's invite and welcome page say an AI interviewer asks the questions and a person reviews the answers, and candidates confirm that before they start. Job seekers can read more in How decisions are made, and the Trust page sets out the rules Emoti is built to keep.

How AI-written text is marked

AI-written text carries a small Emoti mark, such as "Written by Emoti", so the team can tell it apart from what people wrote. Assessments are labeled too: "Assessed by Emoti" on how a profile matches, and "Rated by Emoti against your rubric" on screen answers. Notes stay editable, so a person can correct any part before sharing it.

Which AI models answer

Emoti sends model requests through its own gateway. An admin assigns a model, with fallbacks, to each kind of task, such as reading a brief into criteria, the agent on Home, Ask about a call, and writing notes, and can see which provider and model handled each request. If a model doesn't answer, the next one in line does.

If no model is set up, every feature still works with Emoti's built-in rules, which are simpler: the agent replies from set patterns, a search reads criteria by keyword, and Ask quotes the most relevant parts of the transcript. The page says when they answered, for example "Answered by Emoti's built-in rules". See Troubleshooting.