By the third time in a year that one role family opens, some replies to the recruiter’s first messages say “you contacted me about this in March.” Each search for that family began from the open market, and the people who fit best have now heard from your team more than once. Fewer of them answer, and the team concludes the channel has stopped working.

Point AI sourcing tools only at the open market and they run that search faster, which I’d expect to bring the same people back sooner. So for every sourcing strategy and every tool added to one, I’d ask whether the candidate pool refills between one opening and the next.

Some pools refill from your own hiring: each role that closes leaves finalists who came close, and each person who joins brings former colleagues no one at your company has asked about. Others refill only as the market moves, and I’d expect those to run dry first. I’d build a repeatable candidate pipeline mostly on the first kind, and sort every human resources (HR) sourcing strategy by where its refills come from before deciding where a recruiter’s week goes.

Let every closed role refill the finalist pool.

Your own hiring refills one pool with no sourcing at all. Every role that closes with more than one strong finalist leaves people who know your company and came close after meeting your interviewers, and the next opening in that role family can draw on them.

That pool grows only as fast as roles close. Before a search for a role family, I’d check which roles in the family closed since the last one, and who reached the final stage of each. If none did, the pool hasn’t refilled, and the search moves on to the sources below.

Once a finalist agrees to talk again, deciding which interviews they repeat is walked through in a separate guide.

Ask each new joiner for names while their network is fresh.

Referrals refill from your team’s networks, and a network only brings new names when the person holding it is new to you. Ask a long-serving engineer for names on every engineering role and I’d expect the answers to thin out after the first few asks, because their network hasn’t changed since the last one.

A new joiner’s former colleagues are people no one at your company has asked about. So I’d ask new joiners for names in their first few weeks, while they still remember who was good at what. Ask about a skill the role needs, such as who they’d want beside them on a difficult migration, instead of asking who is good. Then note the role family beside each name, so the next search in that family can find it.

Name one recruiter to own the not-yet replies.

Outbound refills a pool of its own: people who answer a first message about their timing and leave the role itself open, saying they’d talk after a launch or once a promotion round is over. Left in whichever inbox they landed in, those replies go unread when the next search starts.

So name one owner per family for them. That recruiter moves every not-yet reply into the record your team shares with the date your team will write again, and reads the list before each search in the family.

Between those dates, a newsletter-style community for the jobs you fill again and again gives those people something from your team worth opening.

Keep a contact record for every market you search.

The market refills too, as people change jobs, but for a specialist role I’d expect it to refill slowest of all. The map of people who fit can be short enough to read in one sitting, and every message your team sends uses some of it up.

Start a record for each role family’s market, and give every person on it these lines:

• Who they are, and where they work now.
• The date of each message, and the recruiter who sent it.
• The role they were approached about.
• Their reply, pasted in as they wrote it, if one came.

Before the next outbound search for that family, read the record before you read any search results. If most of the map heard from you recently, a faster search reaches the same people again. Widen the map first with adjacent job titles and with companies a size up or down from the ones you searched last time.

Then agree how long someone who didn’t reply waits before your team writes again, and settle it with the recruiters who send the messages. Without that interval written down, the people at the top of every search get two messages from two recruiters in the same month.

Judge AI sourcing tools by the pools they can search.

I count an AI sourcing tool as a way of searching the sources above, so the first thing I’d ask about one is which of those sources it reaches. I’d expect speed on the open market to help least, because the market refills slowest, and a faster search reaches the end of the map sooner.

A search earns its keep on the pools your hiring refills, and those are the hardest to search. Whether a finalist came close shows in what they said in their interviews and what the interviewers wrote down, and an ATS profile holds little of either.

Ask any tool on your shortlist whether it can search the people your recruiters have spoken with before, past interviews included, as well as the open market. Then ask whether every name comes back with a reason a recruiter can check against the brief. Last, ask how its results meet your contact record, because a list that ignores the record sends your team straight back to the people who heard from you in March.

Where Metaview fits in HR sourcing.

The refill check starts with a question about your own interviews: which ones has this role family had since its last search? Metaview Reports takes questions about recorded interviews, so that one can go to the data instead of to whoever remembers.

1

1. 1The “Candidates” header, above five calls listed one per row.

With that answer in hand, give AI Sourcing the role family in plain language, once a year or whenever it opens again. The sourcing agent searches the open web, the interviews your team captured in Metaview, and candidates synced into Metaview from your connected ATS, then lists candidates with its reasoning beside each. In this method, that reasoning gets checked against the contact record: a name the record shows inside its write-again interval waits for its date, however strong the reasoning reads.

Of those sources, interviews are the hardest for a team to read back once a role has closed, the opaque part this line describes.

The Metaview Notetaker records and transcribes a screening call as a visible participant, on calls where the candidate and interviewer have agreed to recording. Metaview captures every spoken word there, so when a candidate says on that call that the timing is wrong, the owner of the not-yet list can set the write-again date from the timing they gave.

Date the last refill of every source.

Pick one role family you hire for every year, and before its next search, write beside each source the date it last refilled. The oldest date tells you which pool to work on first, and a source with no date beside it has no owner yet, so name one that week.