Sourcing for a new role usually goes wrong before the first search runs. The recruiter searches from the job description, so the list fills with everyone who holds the right title. The messages sent to that list read like templates, and the few replies come from people the hiring manager would not interview.
AI talent sourcing gets you that list sooner, and it fails the same way when nothing beside a name says why the person is on it.
I’d judge sourcing for a role by whether the recruiter can write a reason beside every name, in one sentence the hiring manager agrees with. Build the role’s pipeline around those sentences, from the intake call to the first replies, and the same sentence opens each candidate’s first message.
If you’re planning how your team sources across many roles, this guide to talent sourcing strategy covers a plan for each role family.
Turn the intake call into a brief you can search.
On the intake call the hiring manager tells you what the job description left out, so write the brief from the call. A brief that works describes what you would see on a candidate’s profile, which a job description rarely does:
| The job description says | The brief says |
|---|---|
| Excellent communicator | Has presented to customers or to a leadership team, in a role or a talk you can find |
| Experience in a similar role | Has done this job at a company of a similar size, and the brief names a few of those companies |
| Strategic thinker | Has owned a plan from start to finish, such as a launch or a team’s first hires |
Then ask for the dealbreakers: the two or three things that would make the hiring manager pass on a profile that looks right on paper. Write the dealbreakers down on the call, in the hiring manager’s words. They rarely make it into a job description, and they are what the sentence beside each name gets checked against later.
Note where these people work now as well. List the companies and the neighboring job titles the hiring manager would recognize, because a search can aim at those, while a job title on its own matches everyone who holds it. If an AI sourcing tool runs the search, give it the brief with its dealbreakers instead of the job description.
If you’re deciding which channels to search, this guide to recruitment channels covers how to pick them and how many to keep.
Write one sentence for every name you add.
When a candidate goes on the list, write a sentence saying which part of the brief their profile shows and what you would want to ask them.
A sentence that works names something specific on the profile and ties it to the brief, for example, “Led a billing migration at a company the hiring manager named; ask whether she wants to manage a team again.” A sentence like “Strong background, good fit” fails, because the hiring manager has no way to check it.
If you can’t write the sentence, leave the name off. That keeps the list short enough for the hiring manager to read, and every name left on it can be explained.
The rule has a cost. Some strong candidates keep thin profiles, and a sentence is hard to write from the profile alone. For them, write it from the referral or the conversation that brought them to you, and leave the name off only when nothing does.
The same goes for anyone your team has interviewed before, such as a finalist for a similar role. Write their sentence from what they told you in that conversation instead of from their profile, and read how the conversation ended before you add them.
Write each sentence when you add the person, while their profile or notes are still open. Written a day later from memory, it turns into the generic line that fails the test.
Show the sentences to the hiring manager before any outreach.
A brief that reads well can still miss, and the sentences are the quickest way to find out where. So send a calibration batch before a single message goes out: eight to ten names, each with its sentence beside it.
I’d include two or three names you’re unsure about, because the doubtful ones show where the brief is vague. Ask the hiring manager to mark each sentence they agree with and to correct the ones they don’t.
A hiring manager will argue with a sentence far more usefully than with a name. Each correction goes into the brief as a new dealbreaker or a sharper requirement, and if most of the batch comes back corrected, write a second batch from the new brief before you search at scale.
Send the batch within a day or two of the intake call, while the call is still fresh in the hiring manager’s mind.
Open each message with that sentence.
The sentence you wrote for each candidate is the most specific thing you know about them, so put the reason first. Opening on their own work shows the candidate why you wrote to them before they reach the part about the company. I’d build each first message in this order:
- The reason: what you saw on their profile, and why it fits this role.
- The role, and what the team needs from the person who takes it.
- One ask, such as a short call this week, with an easy way to say not now.
- A signature from the recruiter who will run their process.
Plan two follow-ups a few days apart, and give each one something new, such as a line from the hiring manager about the work. A follow-up that only says “bumping this” gives the candidate nothing new to answer.
For the strongest few names, ask the hiring manager to write one line of their own. A line from the person they would work for is hard to mistake for a template, and it takes the hiring manager a minute.
Where Metaview fits in AI talent sourcing.
The slowest part of this method is finding enough names that deserve a sentence. That search is the job of AI Sourcing, and the sentence stays with the recruiter.
AI Sourcing searches the web, your connected applicant tracking system (ATS), and your past Metaview conversations from one plain-language prompt. The sourcing agent returns candidates with the reasoning for each, and it does not contact candidates or assess them on its own.
I’d treat that reasoning as the raw material for the recruiter’s sentence. Check it against the dealbreakers, then write the sentence in words the hiring manager will recognize, or leave the name off.
If results aren’t reliable, AI doesn’t reduce work. It just moves it, and recruiters end up validating, correcting, and second-guessing the output.”
A recruiter who checks each result against the brief before it goes on the list does that correcting once, before any replies arrive.
The search reaches past conversations only if someone recorded them. When everyone on the call consents, the Metaview Notetaker joins the interview as a visible participant and records and transcribes it, then writes structured notes in the template set for that kind of call. Getting that consent is up to your team and its own process.
Metaview captures every spoken word of the interview, so the recruiter who writes to a former finalist months later can start their sentence from what that finalist said they wanted next.
- 1Notes from one screening call under headings such as “Overview” and “Salary expectations,” including that the candidate is motivated by equity.
- 2Templates matched to kinds of call, each with a count of how many times it has been used.
- 3A saved clip from the call, labeled “Motivations for role,” with a “Share Highlight” button beside it.
Those notes can also start a search. In this short demo Siadhal Magos, Metaview’s chief executive, opens his notes from a conversation with a candidate he liked and asks the sourcing agent to find more people like them.
He tells the agent in one line what he liked about the candidate. That line is the sentence this method asks for beside each name, and I’d copy it into the brief as well.
Reports runs over your own interview data, in the product or from an AI assistant you connect to it. Two example reports from an earlier version of its web page show the kind of question a recruiter can put to past interviews before writing to anyone:
- 1The report’s topic, deal size, in account executive interviews only.
- 2Four candidates and the interview each one was in, from a recruiter screen to a final-round chat with the hiring manager.
- 3Each candidate’s answer, including a no from one who focused on smaller deals.
The second example lines up one detail per candidate in its own column.
- 1The column’s name, “Compensation Expectations,” typed in plain words.
- 2Five candidate rows, three with a figure and two still blank.
Read the first replies against the brief.
The first batch of replies says as much about your brief as about the candidates. Four patterns are worth reading for, and each one points at a different line of the brief:
| What the replies show | Change before the next batch |
|---|---|
| People say the role is too junior or too senior for them | The level in the brief: ask the hiring manager which way it should move, and rewrite the sentences that leaned on it |
| Keen replies from candidates the hiring manager then turns down | A dealbreaker the brief never captured: ask what it was and add it |
| No replies at all from the companies you aimed at | The target list: move the search to the neighboring titles the brief names, and ask the hiring manager for other companies |
| Polite replies saying “not now” | Nothing in the brief: note when the candidate asked to hear from you again, and put that date on their record in your ATS |
Before the next batch, I’d take every sentence the hiring manager corrected and fix the line of the brief behind it, then write the new names’ sentences against that version. A batch the hiring manager reads without correcting a single sentence is a pipeline you can defend, one name at a time, and that is the point to search at full volume.
Build a pipeline you can explain name by name.
See how AI Sourcing puts its reasoning beside every candidate it returns, so each sentence on your list starts from something the recruiter can check.
Frequently asked.
How many candidates should you source for one role?
Keep adding names while the recruiter can still write a strong reason for the next one, then stop and wait for replies. For a hard role I’d send a first batch of 20 to 30 messages and widen the search only after those come back.
What if the hiring manager can’t name any dealbreakers?
Show them two profiles you think they would reject and ask what is wrong with each. Most hiring managers find it easier to say no to a real person than to describe one in the abstract, and their reasons are the dealbreakers.
Should the sentence beside each name go in your ATS?
Yes, as a note on the candidate’s record, so whoever picks up the role later can see why each person was approached. Write it as you would be comfortable with the candidate reading it, because privacy law in many places lets candidates ask to see what you hold on them.
How does this work for a confidential search?
The sentence is about the candidate, so it works the same way. Keep the company’s name out of the first message and describe the team and the problem instead. Name the company once the candidate agrees to a call.
Does a sentence per name slow down high-volume hiring?
Use it for the people your team approaches. Inbound applicants chose the role themselves, so the sentence is worth its time where your team starts the conversation.
