Recruitment: Three Loops, One Pool
Hiring is not a straight pipeline, because real hiring never is. Three loops run all the time: demand (roles open, pause, reopen, and close), sourcing (always on, across every channel), and selection (screen, interview as many rounds as it takes, then offer). All three feed one candidate pool that never forgets anyone. People move backward as often as forward, and every way out of a loop is a way into the pool.
Hiring is drawn as a line and lived as a loop
Most hiring processes fail in the gap between the tidy diagram and the messy reality:
Three loops orbiting one pool
Each loop runs on its own clock. None of them waits for the others, and all of them read from and write to the same candidate pool.
Roles keep changing, the system keeps up
A recruiter opens the role from the Job Reqs list: title, employment type, location, remote policy, salary band. The system sets up the same five-stage board every role uses (Screen, Interview, Offer, Hired, Rejected), so no two roles run slightly different processes. The role is open for hiring right away but stays off the public site until the posting is ready.
The public posting is written on the role itself: the job description, a clean URL, and up to three screening questions. Making it public puts it on the careers page. Some roles are never posted at all and get filled from sourcing and the pool. The posting is one door among several, not the whole process.
Budgets move and priorities shift, and the role you start hiring for is not always the role you finish hiring for. A role can go on hold and come back, and its description, salary band, and screening questions can be edited mid-search. Candidates already in it keep their full history through every change. Nothing resets just because the role changed.
Closing a role records an outcome (filled, closed without a hire, or cancelled) and takes it off the careers page automatically. The candidates in it do not disappear: they move into the pool with their screens, ratings, and notes attached, and the strong ones show up first when a similar role opens. Reopening a role picks up right where it left off.
Always on, across every channel
A candidate applies with a resume, cover letter, and answers to the role’s screening questions. The application lands in the talent system attached to the role. No inbox, no forwarding, no resume that lives only in one person’s email.
LinkedIn sourcing, team referrals, event contacts, and agency submissions all run alongside inbound applications. Sourcing is always on, not a burst that starts when a role opens. Every channel is tagged when the candidate comes in (sourced, referral, agency, LinkedIn, job board, event), so the system can later tell which channels actually produce hires.
Agency candidates come in through the same doors and get the same treatment as everyone else: the same AI screen, the same two gates, the same record. No separate spreadsheet for agency submissions. One record no matter who found the person.
Sourced resumes come in batches: a recruiter drops up to 25 files at once and the AI reads each one, filling in a draft with name, email, phone, LinkedIn, headline, and current title. The recruiter checks and saves each one. Ten minutes of drag-and-drop replaces an afternoon of typing, and duplicates can’t happen: people are matched by email, one application per person per role, so re-adding someone opens their existing record instead of making a second.
When a role opens, sourcing starts from everyone the company has already met. The pool is ranked by best AI screen and grouped by role family, so last quarter’s strong runner-up shows up at the top of this quarter’s search with full history attached. The easiest candidate to find is the one you already found.
The real flowchart, every branch included
One candidate’s path through selection. Solid lines are the working process; dashed lines are the pool’s memory. Every “no” leads somewhere, and none of them is a shredder.
AI on the interview panel
- The AI holds a seat on every round, from the recruiter screen to the founder chat, and a round with three people hands in four scorecards
- It scores against that round’s rubric with an advance, hold, or reject call, citing the exact lines and timestamps it used
- Its scorecard stays sealed until every person on the round has submitted, so it can never anchor the room
- It carries memory forward: anything a transcript left unproven becomes a question for the next round
- Rough transcription is handled, not trusted: it flags low-confidence quotes and never holds a garbled transcript against the candidate
Human and AI, side by side
- Two separate reads at the resume, and again at every interview, catch what one reader alone would miss
- The AI gives its read with evidence; the people make the decision
- When two scores are a full point apart, that gap is raised in the debrief, not averaged away
- No one moves forward, or gets cut, on an AI score alone
The moment an application lands, from any door, Claude reads the whole resume against the job description and writes a structured screen: a 0 to 5 fit rating, a short overview, clear strengths and gaps, a read on English, and the salary expectation and notice period exactly as stated, never guessed. Every application gets the same careful read, whether it came in first or five hundredth.
Scored candidates are also ranked within their role family, not just the one job they applied to. A strong engineer who applied to the wrong opening still shows up near the top of the engineering family.
Every application shows the AI’s score with its written reasoning next to the recruiter’s own star rating. Weak on both: rejected, with the reason recorded, into the pool. Strong on both: on to the screening call. The two disagree: a second person takes a look and decides, because disagreement is a signal, not noise. No one is rejected by the AI alone.
Before any formal interview, a recruiter talks to the candidate: motivation, expectations, notice period, the things a resume can’t show. Right role at the right time: on to interviews. Wrong role or wrong time: parked as future consideration, or marked as withdrew if the candidate steps back. Either way the call notes go on the record, and the pool keeps them.
The number of interviews is not fixed. A senior role may take three rounds plus a follow-up with a different interviewer; a junior role may take one. Panel says yes: offer. Panel says no: rejected, with the reason recorded. Panel is split: another round, and that step back is recorded like any other. No off-the-record “can you talk to them once more” that the system never sees.
The AI sits on every interview panel too. It reads that transcript against the job description, the resume, the earlier AI screen, the company’s core values, and every prior round, then fills in a scorecard: an advance, hold, or reject call, a score for each thing the round is judged on, and the exact quotes and timestamps it based them on. Where the transcription is rough, it flags the low-confidence lines and never holds a garbled transcript against the candidate. Its scorecard stays hidden until the human interviewers hand in theirs, so it can’t sway the room. It also remembers the earlier rounds, so anything left unproven in one round becomes a question for the next. The AI is one voice on the panel. It never makes the final call.
Everything about the candidate lives on the application: the notes, the resume (replaceable when a better version arrives), the cover letter and answers, every scorecard from people and AI, and the stage history. Anyone on the team can open it and know exactly where things stand.
Offers get negotiated, and terms move both ways before they settle. Accepted: the application flips to hired and hands off to the New Member Onboarding workflow, which turns the applicant record into an employee record without retyping anything. Declined: the search goes back to the warm shortlist, still ranked and still in the system, not back to square one.
Either way, the loop closes with a clear status and a recorded reason. No one is left in limbo, and no outcome is silent.
Every exit is a pool entry
“Rejected” is a status, not a delete. Every way out of the three loops lands in the pool with full history:
The ways out
- Rejected, always with a recorded reason
- Withdrew: candidates change their minds; the door stays open
- Future consideration: right person, wrong timing, parked deliberately
- On hold / passive: still going but paused, usually with the role
- Hired: off to onboarding, still on the record
What the pool does with them
- Ranks everyone by their best AI screen, across every application
- Groups by role family so the next search starts pre-sorted
- Keeps every screen, rating, and note attached to the person
- Feeds the sourcing loop: resurfaced candidates skip the cold start
- Honors a do-not-hire flag where the decision is final
The seven elements
Every workflow we document has the same anatomy: seven elements, each assigned to a human, a machine, or both. This is the Centaur Map from our workflow design method.
No single trigger. A role opening, a strong inbound resume, a referral, or someone resurfacing from the pool can each start things off, and usually several are running at once.
The job description and screening questions, resumes from every channel (careers page, LinkedIn, referrals, agencies, batch drops), and the full history of everyone the company has already met.
Every candidate gets two reads at the resume: the AI screen and the recruiter’s rating. The AI then joins every interview too, scoring on its own before it sees what the people said. People make every call to advance, reject, or send someone back.
Not forward-only. Candidates move back a stage for another round, return to the shortlist after a declined offer, or leave to the pool and come back months later on a different role.
A living record for each candidate: every application, every screen, both ratings, the notes, every interview scorecard, and every status change with its reason.
Ranked, sortable views wherever the work happens: by role, by role family, and across the whole pool.
How many make it through each loop, how often the AI and the recruiter disagree, the time from opening a role to hiring, and how often the pool, not a job board, fills the role.
The standing rules
- Every role runs the same five-stage board. The loops vary; the record doesn’t
- Every door produces the same structured record, agency or inbound alike
- The AI reads every resume in full as it arrives
- Backward is a normal direction, and every move is on the record
- The AI is on every interview panel and scores before it sees the people’s scores; it advises, it never decides
- No candidate is rejected by the AI alone, and every exit has a recorded reason
- Closing a role never throws away its candidates; nothing is ever deleted
Why it works
- The written process matches the real one, so people actually keep the system up to date
- One record ends the “where does this candidate stand” question
- Always-on sourcing means a new role starts warm, not cold
- Two separate gates catch what either one alone would miss
- Every interview is written down with its evidence and carried forward, so three rounds add up to one story instead of three separate opinions
- The pool builds up: every search makes the next one faster
- Declined offers and paused roles cost days, not restarts