Every candidate gets the same structured screen — conducted by the AI voice interviewer, or by your own recruiter while the AI stays silent and scores. Out comes an evidence-backed scorecard, quoted from the transcript. Agencies send it to their client; in-house teams decide with it.
The screen you ran today answers the next requirement — your client's, or your own.
The interview, the scorecard and the pool are identical whether you are screening for a client or for yourself. What changes is who reads the scorecard at the end — so that is the only place this page splits.
Multi-client. Your job is to put a defensible shortlist in front of someone else, faster than the agency down the road.
Overlapping bookings are fine — the AI runs rooms in parallel, so round one is not rationed by who is free to sit in it.
A client-facing scorecard link carries the scores, the quoted moments, the coding and MCQ results and the recording. Links are off until you turn one on, and revocable after.
A no for this role is not wasted work. The profile and its scorecard stay in your pool and surface against every future requirement, whoever the client is.
The resume scanner reads a CV against the JD and scores the fit, so the interview calendar goes to the people worth interviewing.
The scorecard is your handoff — it has to survive a client reading it.
One company, your own roles. No handoff, no shortlist to sell: you read the scorecard and decide who reaches your team.
A hosted careers page and Indeed intake drop applications straight onto the role, so nobody re-types a resume into a spreadsheet.
Your roles carry their own screening bar and question set, so applicant number forty is asked at the same standard as number one.
Per-skill scores with the transcript line behind each one — the thing a hiring manager can actually argue with, instead of "seemed strong".
Everyone you have screened stays searchable by skill and score across your roles, and the resume database holds the people you scanned but have not interviewed yet — so the next opening does not start from an empty queue.
The scorecard is your decision record — nobody downstream to convince.
Chosen per booking. Recruiter-led means a human asks the questions and the AI stays silent — and still scores the answers.
A recruiter or panel can walk into a running AI interview, mute the AI and ask their own questions. Not an async video you watch later.
The scorecard reports what the screen actually covered — spoken answers, MCQ, coding, recording — and only scores the sections that ran.
When there is not enough verified evidence to stand behind a read, the scorecard says so instead of inventing a verdict.
Most screening work is thrown away the moment the role closes. SourceLoop keeps it — the scores, the quoted evidence, the skills — so the next requirement starts from people you have already assessed. Your client's next brief, or your own next opening.
Applicants in, screen run, evidence out — and the screen still working months later. The same nine pieces whether the shortlist goes to a client or straight to your own hiring manager.
A hosted careers page and Indeed intake drop applications straight onto the role — resume attached, nothing re-typed. In-house teams screen their own applicants; agencies work inbound and sourced candidates from the same queue.
The scanner reads a CV against the JD, scores the fit and files it into your candidate database — so interview slots go to the people worth interviewing, and the rest are still searchable later.
Pick AI-led or recruiter-led on every booking. Recruiter-led means a human asks the questions and the AI stays silent — and still scores the answers. Overlapping slots are fine: the AI runs rooms in parallel, and invites plus ICS files go out on their own.
A voice interviewer works through questions built from the resume and the JD, then MCQs and a coding problem in the same room — at 3am if that is when your candidate is free. No panel to coordinate.
A recruiter or panel can join a running AI interview, mute the AI and ask their own questions — then hand it back. Everything said is still transcribed and scored. Async one-way interviews cannot do this.
Per-skill scores, each one carrying the verbatim transcript line it came from. Quotes are checked back against the transcript before they count — and when too few survive, the scorecard is reported as inconclusive rather than guessing a verdict.
The scorecard says what the screen covered — spoken answers, MCQ, coding, recording — and only scores the sections that ran. A round that was never administered reads as not assessed, not as a low score.
One link carries the recording, transcript, MCQ and coding work and the per-skill scores. It is off until you switch it on, and revocable after — an agency sends it to the client, an in-house team sends it to the hiring manager.
The profile and its scorecard stay searchable by skill and minimum score. On an agency desk that search re-matches the same person against every client requirement you open after; in-house it runs across your own roles, so a second opening starts from applicants you have already assessed. Nobody gets screened twice for the same work.
Not just the interview — the desk around it. The bar is set once and locked, then the agency lane and the in-house lane run their own workflow, and both arrive at the same evidence.
“We cut the write path over first and left the old index read-only for a week, so a rollback was one config flip.”
The scorecard the app renders: per-skill scores with the transcript line each one came from, the scorer's own confidence, and what the screen actually covered. When the evidence will not carry a verdict, it says inconclusive.
Set the bar once. Run it every day — open the full app
Agency or in-house, first-round screening is the same tax: senior hours, calendar ping-pong, and a shortlist built on notes nobody else can audit. Six ways it gets paid.
Panel calendars, time zones, reschedules — a screen that could run tonight happens next Thursday. If the candidate hasn't gone somewhere else by then.
Every first-round screen spends an engineer or a senior recruiter. Headcount caps how many you can run, and that cap sets everything downstream.
Two interviewers, two sets of questions, two standards. Applicant forty is not measured against the same thing as applicant one, and nobody can tell you which was harder.
"Seemed strong on Kubernetes" is not something a client can check. One under-screened candidate in front of them undoes ten good sends — your shortlist is your reputation.
Careers-page and job-board applications arrive faster than anyone can read them, so good applicants time out in a queue while the team screens whoever floated to the top.
The person you assessed last quarter is a stranger again next quarter. No scores, no evidence, no notes worth re-reading — so the next role starts from a blank page.
Every screen you run from today keeps its scores and its evidence — ready for your client's next requirement, or your own next opening.
Staffing & sourcing agencies · in-house hiring teams · AI-led or recruiter-led, your call per interview