# 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