How to Write a Hiring Brief for AI Sourcing (The Intake That Fixes Your Pipeline)

A hiring brief for AI sourcing is a short structured document that states the problem the hire solves, the signals in a profile that predict success in your environment, and the reasons you would pass on someone who looks right on paper. Write those three things and your candidate quality improves immediately, on any tool. Skip them and no amount of model quality will save the search.

If your AI sourcing is returning people who are technically on target and practically useless, the model is almost never the culprit. The brief is.

Your intake notes were written for people who already know the context

Most hiring intake is a set of shorthand notes taken by a recruiter who has worked with that hiring manager for two years. "Same as the last one, but more senior, must be comfortable with ambiguity." A human recruiter can run on that. A sourcing system cannot, because it has no shared history with your team. It has your words and nothing else.

So the brief has to carry the context that used to live in a relationship. That is more writing up front, and it is the highest-leverage 45 minutes in the entire search.

Start with the problem, not the title

Open the brief with two or three sentences describing what will be true in twelve months if this hire goes well. Not responsibilities. Outcomes.

Weak: "Senior Data Engineer to join our platform team."

Strong: "Our reporting pipeline breaks weekly and nobody owns it. In twelve months this person has rebuilt ingestion, cut the on-call load, and the analytics team stops filing tickets to get numbers."

The second version gives you search terms, screening questions, and outreach copy for free. It also tells you something the title never would: the people who thrive here have lived through the same mess somewhere else. That is the whole basis of finding talent that isn't looking, which we covered in why the best candidates never apply.

List five people you would hire tomorrow, and say why

Before any tool touches the role, find five real profiles by hand and write one sentence each explaining what makes them interesting.

Examples of good sentences:

• "Ran the same migration at a company one size up from ours."
• "Came out of a team known for this stack, so the ramp is short."
• "Two years into a role that usually turns over at three."

That list does three jobs. It surfaces your real criteria, which are rarely the ones in the job description. It becomes a calibration set, so you can ask any vendor or any internal build whether it would have surfaced these five. And it gives the hiring manager something concrete to react to, which is a far better use of a meeting than another round of adjectives.

Write your rejection criteria, the section everyone skips

Every brief lists what you want. Almost none list what you will pass on. This is the single biggest reason AI sourcing feels noisy, because a system optimizing only for match has no idea what a false positive looks like to you.

Be blunt and be specific:

• "Agency background only, with no in-house experience, is a no for this role."
• "Five roles in six years reads as a risk here given the project length."
• "Big-company pedigree without hands-on work is the exact failure mode we hired into last time."

Write three to five of these. If your hiring manager balks at putting them in writing, that is useful information: it means the bar is not agreed on yet, and no tool is going to resolve that for you.

Define signals, not keywords

Keywords are what a profile says. Signals are what a profile implies. A brief that only carries keywords produces a pipeline of people who wrote a good LinkedIn headline.

For each requirement, write the observable signal underneath it:

• Requirement: comfortable with ambiguity. Signal: joined a team before the function existed, or was employee number one in a role.
• Requirement: strong ownership. Signal: shipped something that is still in production and named in their profile.
• Requirement: scale experience. Signal: worked at a company between 200 and 2,000 people, because that is the range where our problems actually show up.

This section is what makes a brief portable. It survives a tooling change, and it is the part your team will reuse across five similar searches. It also lines up with the judgement calls a model cannot make for you, which we broke down in what not to automate in recruiting.

Say what the outreach is allowed to promise

Sourcing and outreach are one system. If the brief does not include the pitch, someone writes the pitch under time pressure, and it comes out generic.

Include four things: why this role is genuinely interesting to a person who is not looking, what is hard about it (say it out loud, strong candidates trust the honesty), the compensation range you can actually approve, and the two or three things about the team you would say on a call.

A model can write a personalized message in seconds. It cannot invent a reason for someone to leave a good job. That reason has to come from the brief. If you want to know whether the messages going out are actually working, the metrics that tell you cover it.

Put a decision rule at the bottom

One line: what happens when a candidate matches the signals but misses a stated requirement.

"If they hit the pipeline-ownership signal but lack the specific tool experience, we screen them." or "No exceptions on the certification, it is a client requirement."

This is the line that keeps the search from stalling in week three, and it is the line that tells whoever is running your sourcing how much latitude they have.

Re-run the brief after the first ten profiles

Treat the brief as version one. After the first batch of candidates, sit with the hiring manager, sort them into yes and no, and ask why for each one. Then update the brief in the same sitting.

Most teams do this reactively through weeks of feedback in a chat thread. Doing it once, deliberately, in a 20-minute review cuts the calibration cycle dramatically, because you are correcting the input rather than correcting the output over and over.

The brief in one page

Copy this structure:

1. Outcome in twelve months, three sentences.
2. Five real profiles, one sentence each on why.
3. Signals, one line under each requirement.
4. Rejection criteria, three to five, written plainly.
5. What outreach can promise, including range.
6. Decision rule for the edge case.
7. Version and review date.

If it runs longer than a page, you are writing a job description, not a brief.

Where this leaves the tooling question

Good AI sourcing is real and it is a large advantage. It is also downstream of judgment. The teams getting value out of it are not the ones with the best vendor, they are the ones who did the thinking that makes any vendor work.

If you want a partner doing this alongside your team, the AI Talent Partner program at nextchaptertalent.ai builds the structured brief at the front of every search and runs sourcing and outreach off it. And if you landed here from the candidate side, the free tools at nextchaptertalent.com/templates and the program details at nextchaptertalent.com/pricing are the place to start.

FAQ

How long should a hiring brief be?

One page. Long enough to carry the outcome, the signals, and the rejection criteria, short enough that a hiring manager will actually read it before the kickoff call. Anything longer usually means you copied the job description into it.

What is the difference between a hiring brief and a job description?

A job description is external marketing written for applicants. A hiring brief is internal calibration written for whoever is sourcing, screening, and deciding. The brief includes things you would never publish, like your rejection criteria and the honest hard parts of the role.

Do I need to rewrite the brief for every role?

No. The signals section is highly reusable across similar roles, so your second search on the same team should take half the time. What changes every time is the outcome statement and the rejection criteria, because those are specific to the gap you are filling right now.