If your talent sourcing strategy is a list of channels, every open role gets the same treatment from it. When a role opens, the recruiter posts it and messages strangers, whether your team hires for that role every month or once every few years. At the end of the quarter, it’s hard to say which of those sources found the people you hired.

I’d find the strongest candidates for a role the way I’d expect an expert recruiter to: from a written plan for that kind of role. If artificial intelligence (AI) software runs some of your searches, your AI sourcing strategy is the part of the plan that says what the software searches first. Write the plan down as your talent sourcing strategy, a role family at a time. Each family’s plan says where the search starts and who does each part of the work, and once a quarter you check it against where that family’s hires came from.

For a family you hire often, the people who reached its final round last time have already sat through your interviews, so I’d start with people you know there and go to the open market second.

Sort your roles by how often you hire them.

Start by deciding which roles share a plan. Group your open and expected roles into families, then ask two questions of each family: how often you open a role in it, and how many qualified people the market holds for it.

The answers put most families into one of four kinds, and each kind starts its search somewhere different:

Keep the families broad enough that each one has a role open most quarters, or the plan goes stale before anyone uses it. Account executives across two segments can share a plan. A first head of finance can’t share one with anybody.

I’d write plans for repeat families first, because they are where the same search gets run from scratch most often. A role you open once every two years can live on a short note until it opens.

Start each frequent family with people your team already knows.

I think the order of sources saves more hours than any other part of the plan, because it decides how much of each search goes to people you’ve never met. When a family is one you hire often, I’d put your own past finalists first and the open market last:

1. Finalists and silver medalists from past searches for the same family, who already know your interviewers and your process.
2. People your team interviewed for a neighboring role, whose notes say what they were looking for when you met.
3. Referrals, where you send employees the written brief and ask who fits it.
4. Employees who could move into the role, once you’ve checked with their manager and your people team.
5. Profile searches on the open market, and the communities where people in that family gather. Niche community sourcing and personalized cold outreach are two candidate sourcing strategies that fit this step.

The first two come from your applicant tracking system (ATS) and your interview notes, and they’re easy to skip, because the open-market search is the habit. I’d expect a note from a recruiter a candidate has already met to be easier to answer than a cold message, and you already know what they said about the work they want next. So search past finalists first, before any outreach to strangers goes out.

Which outside channels earn a place in step five is a separate decision, made once for the whole team. Choosing which recruitment channels to keep is covered in a post of its own, including when to drop one.

Decide what your AI sourcing strategy hands to software.

Once software can run a search from a written description of the role, the plan has to say which parts of the work it takes and which stay with people. I’d split each search between the software and the people who own the role.

Give the software the repeated searching, starting with your own past candidates and running it again whenever the brief changes. Keep the brief and every message with the recruiter, who reads why each candidate was suggested before writing to anyone. Have the hiring manager review the first batch and say what is missing before the search runs at full volume.

I’d put the hiring manager’s review that early because their first reaction to real candidates should tell you more about the role than the job description did, and it’s cheaper to hear it before the recruiter starts writing to people.

I’d also make sure a recruiter writes every message. A past finalist has already given your team hours of interviews, so a careless note to them costs more than rerunning a search. For the note itself, there’s a walkthrough on personalizing recruiting emails at scale.

Write each plan down at the intake call.

A plan that lives in the recruiter’s head can’t be reviewed, so write it down and agree on it with the hiring manager before the intake call ends. One page per family is enough, and it needs five things:

• The family and the roles it covers.
• Three or four must-haves the hiring manager has agreed to, in words a recruiter can search for.
• The source order for this family, starting with the candidates you already know.
• Who runs each part of the search, and when the hiring manager sees the first candidates.
• What you’ll check at the next review, and the date of that review.

The written order also settles the “more candidates” request. When a hiring manager asks for more, you can show which sources have run and which come next, and agree whether to move down the list or change the brief.

Where Metaview fits in a talent sourcing strategy.

A plan for a rare senior or specialist family can’t name the companies its search starts from until someone has read the market. Deep Research, a mode of Metaview’s AI Sourcing agent, is meant for that market read.

Ask it the question the plan’s starting point depends on, for instance where a named company has been hiring engineers from over the past year. Deep Research answers in a report or a set of profiles, and the recruiter picks the companies the family’s search starts from.

For a frequent family, write the brief as one natural-language prompt, and the same agent searches your connected ATS and your team’s past Metaview conversations as well as the web. It hands back a list of candidates, each with the reason it was chosen. Messages and fit decisions stay with your recruiters, and the reasoning beside each name is there for them to check against the brief.

Your own interview data can sharpen the first step of a frequent family’s order. Metaview Reports lets you query your team’s recorded interviews, and for an account executive family I’d ask which past candidates had discussed large deals. Metaview has illustrated that kind of report with a column headed “Discussed deals > $100K”:

1

2

1. 1Which interview produced each answer, such as a “Hiring Manager Interview.”
2. 2Each candidate’s answer under “Discussed deals > $100K,” with a short note on what they said.

The Metaview Notetaker is where those answers come from. On an interview the candidate agreed to have recorded, it comes in as a visible participant and leaves structured notes of the conversation behind.

Metaview captures every spoken word, so a past finalist’s answers to the questions behind the family’s must-haves are there to read when a role reopens, and a recruiter can decide who to approach first from them.

1

2

1. 1“Motivated by equity,” written under “Salary expectations,” which tells a recruiter what to lead with if this candidate hears about a new role.
2. 2A rule that applies one note template to every recruiter screen for an “Account Exec” role, so each screen in that family is written up the same way.

Pay is in the same record. Another Metaview illustration shows a column for the pay each candidate expected, worth reading before a past candidate hears about a new role:

1

2

1. 1A new column header asking for each person’s pay expectations.
2. 2Gray bars on two rows the column has not reached yet.

Searching what past candidates said in their interviews, alongside their profiles, is close to a point Siadhal Magos made in a post about AI screening:

Review every plan against its last hires.

Once a quarter, put each family’s plan next to what happened, with your ATS records for the quarter open beside it. I’d check which step of the written order each finalist and hire came from.

In a frequent family, past finalists stay first. If a later step produced most of the rest, move it up among the sources that follow them, and if almost none came from your own past candidates, check whether the recruiter searched them before going to the open market. For a rare family, the market read stays first, and the check is whether the search started from the companies that read named.

Your ATS holds the source of each hire if recruiters record it. A separate guide on tracking source of hire covers how to record it, so that check has something to read.

I’d hold any expert recruiter to that quarterly check, and it’s how I’d run sourcing to find the strongest candidates for a role: past finalists first and the open market last for a family you hire often, and a read of the market first for one you rarely hire. After the first review, I’d rewrite each family’s source order below its fixed first step, and put the date of the next review on the same page.