AIN People/AIN Hire

Four assistants.
One hire you can explain

One sentence becomes a published role. Every CV that arrives is read against the weights you set. The strongest interview on their own phone. Every number opens to the line behind it — and every decision stays yours.

Watch the film · 1:41 Open AIN Hire
JD · Screening & Interview · Report · Talent Bahasa Indonesia native Decision support — humans decide
AIN Hire — your AI-powered hiring assistant 1:41 · One position, 481 applicants, four assistants

Filmed inside our showcase workspace. The applicants are synthetic and labelled as such; the screens, the scores, and the arithmetic behind them are the product's own.

One sentence → a live role
Type “Customer Service Officer, Bandung branch.” AIN drafts the description, the skills, the prerequisites, and every stage an applicant passes through. Nothing publishes until you approve it.
481 CVs, all read
Not a sample. Every application is scored component by component against the weights you set before the posting went out.
Every number opens
Behind each score is a quoted line from the candidate's own CV or interview, with the skill it maps to and the limits of the evidence.
AI-powered, human-decided
The assistants read, score, and write. Shortlist, reject, hire, promote — all human, all with a recorded reason.
The product

Four assistants,
named for what they do

When one posting draws thousands of applicants, screening becomes triage. Each assistant takes something you already write and returns something you already need — and shows its work on the way.

Hiring today
  • Recruiters read a fraction of the CVs that arrive.
  • Criteria shift between reviewers, and between weeks.
  • A gap in a CV quietly becomes a low score.
  • Six months later, nobody can reconstruct why a candidate was cut.
Hiring with AIN Hire
  • Every applicant is read against the same published criteria.
  • Criteria are frozen per version, so a round stays comparable end to end.
  • Missing evidence is flagged for review, never scored as a zero.
  • Every score keeps its quote, and every decision keeps its reason.
Assistant one
JD Assistant
You describe the role in a sentence. AIN drafts the whole hiring instrument — and it can't go live until a human signs it off.
You give
“Customer Service Officer, Bandung branch.”
You get
Job description · skills · prerequisite questions · the full applicant pipeline — as an editable draft
  • Weight each competency yourself — the showcase role runs Customer empathy 30, Communication 25, Problem solving 20, Process discipline 15, Data literacy 10.
  • Those numbers bind every judgement that follows, and they are fixed before the first CV arrives.
  • Draft → review → approval → publish, every transition audited. Publishing produces a version-pinned job URL and a printable QR, so an edit later never invalidates a printed poster.
Assistant two
Screening & Interview Assistant
The part that doesn't scale by hand: reading thousands of CVs, then interviewing the strongest without booking a single room.
You give
Five interview questions, each mapped to the skill it tests (plus an optional written assessment)
You get
Every applicant scored, ranked, and quotable — CV and interview combined under your weights
  • 481 applicants, 481 CVs read. Four components — experience fit, skill match, domain fit, impact evidence — each with its own weight.
  • Candidates interview on their own phone, in Bahasa Indonesia, whenever they're ready. Answers are transcribed server-side and scored against anchored levels, never impressions.
  • Missing evidence is a review state, not a zero — in the showcase round, 455 screened with full evidence, 26 flagged for a recruiter, 0 scored without a source.
Assistant three
Report Assistant
The funnel, the drop-offs, and the evidence coverage — asked for in plain words and answered in the shape the meeting needs.
You give
“Summarise this month's funnel and where candidates drop out.”
You get
A written analysis, generated server-side from the same numbers on your screen — as PDF, slides, or a web page
  • The showcase funnel, live: Application 481 → CV screening 481 → Interview 38 → Decision 9.
  • It names the steepest fall rather than making you find it — “the steepest fall is CV screening to interview, where 443 stop.”
  • Hiring funnel, evidence coverage, score distribution, and decisions — one document, ready to send.
Assistant four
Talent Assistant
Every strong candidate you didn't hire is still evidence. Ask your talent pool a question and it answers against the same rules.
You give
“Who else fits a Relationship Officer in Bandung?”
You get
Ranked candidates from the pool, each with the evidence behind the match
  • Wrong fit for one role is often a good fit for another — Lukman surfaces at 55 for Relationship Officer, Bandung and 50 for Telesales Officer, Jakarta.
  • Same fairness rule everywhere: a stage that has not run is not counted as zero.
  • Works for internal mobility too — employees nominated from the pool are reported as ready now, ready with support, development needed, or evidence insufficient.
One hire, end to end

A 97 that doesn't
survive the interview

Lukman Sinaga's CV came second out of 481. His interview said something else — and the platform shows you exactly where the number moved, and why.

Step 1 · CV–Role Match
97
Four components, your weights, read straight from the CV.
Experience fit×3096
Skill match×3098
Domain fit×2095
Impact evidence×2099
Second of 481 on the CV alone. “Lihat bukti” on the impact score opens the quoted line: CV · Work experience, line 4.
Step 2 · Structured interview
55
Recorded on his own phone, transcribed server-side, scored per question–skill mapping.
Customer empathy — level 3/5×403
Communication — level 3/5×353
Problem solving — level 2/5×252
Anchored levels, not impressions. Every level keeps the verbatim quote it was awarded for, and weak audio goes to a human reviewer instead of scoring low.
Step 3 · Combined
69
97 at 33% + 55 at 67% — the stage weights the company set, applied the same way to all 481.
CV evidence33%97
Interview evidence67%55
Strong on paper. Not for this role — and you can put that sentence in front of the candidate, the hiring manager, or an auditor with the evidence attached.
“Not recommended to advance” — written by a recruiter, not by the model
The platform produces the number, the components, and the quotes. A person reads them and makes the call, with a reason that becomes part of an append-only record. Six months later that record still reconstructs the decision — including the fact that the CV was excellent.
And he doesn't disappear
Relationship Officer · Bandung55
Telesales Officer · Jakarta50
The Talent Assistant carries the same evidence into other open roles, so a no here isn't a dead end for a candidate you spent 481 CVs finding.
What's real in this example: Lukman is a synthetic candidate in our showcase workspace, and we label him as one. His CV score of 97, his rank, the quoted evidence line, and the “not recommended to advance” outcome are actual rows in the product. The interview breakdown — which competencies carried which weight, and the anchored levels behind the 55 — is an illustrative reconstruction, chosen so the arithmetic checks out end to end. On your own position, you'd see your real breakdown instead.
Both sides of the hire

A phone for the candidate,
a workspace for the recruiter

The candidate journey runs entirely on a phone browser, in Bahasa Indonesia. The recruiter side does the reading, the transcribing, and the bookkeeping — and hands every judgement back to a person.

The candidate
Apply first, prove it in the flow
No app, no account wall, and a real alternative path when the standard route doesn't fit.
  1. Open the job. Every published position gets a public page with its own URL and a printable QR, pinned to the version it was published under.
  2. Apply in one pass. Profile, prerequisite answers, purpose-specific consent, and a private CV upload — one form, no account creation in front of it.
  3. Interview on your own phone. Questions written by the recruiters, answered in turns of up to sixty seconds, transcribed server-side and scored against your weights.
  4. Ask for another route. When the device, the connection, or an accessibility need makes that unsuitable, request a phone interview, a written response, or an assisted session — resolved by a recruiter with a recorded reason.
  5. Check status any time. A six-digit email code — hashed, single-use, rate-limited, ten-minute expiry — opens a status page. Stage labels only.
Consent is recorded per purpose, not as one blanket tick. Video is separate from audio and optional — declining it leaves a genuine audio-only path, not a degraded one. Candidates never see scores, rank, or cross-employer data.
The recruiter
Built for thousands of applicants
Recruiters define the role, approve what publishes, and make every call that affects a candidate.
  1. Position & role blueprint. A position owns its job description, criteria, and applicants. Blueprint versions are immutable; the next round is a duplicate with a clean applicant set.
  2. AI drafts, humans approve. Drafting is AI; draft → review → approval → publish is not, and every transition is audited.
  3. Top-N gate, with a reason. Preview and confirm the top 100, 500, 1,000, 2,000, or a custom cut. The confirmation is audited and advances a stage — it never rejects anyone on its own.
  4. Interview kit. Author questions, map each to the skills it tests, set the mapping weight. Server-side transcription produces the scoring text; a browser note can't bypass it.
  5. Review queue for weak audio. If transcription returns nothing usable, a reviewer replays, retries, or adds a correction as a new revision. Scoring resumes only once the text is real.
  6. Shortlist, handoff, sign-off. Decisions are individual and carry reasons. Shortlisted candidates reach the hiring manager through a recipient-bound, expiring, one-time link.
CV–Role Match runs four editable components, with trajectory and stability reported as a factual flag rather than folded into the score. Internal mobility uses the same evidence rules, reported in readiness language instead of pass or fail.
Responsible AI

Defensible by design,
not by promise

Hiring decisions get challenged. AIN Hire is built so you can show exactly how one was reached — and so the AI never makes the final call.

Decision support, not decision making
AI extracts and organizes job-related evidence. It cannot make a final shortlist, rejection, hire, promotion, or rotation decision. Human decisions and overrides require reasons and create append-only revisions.
Consent, versioned per purpose
Consent is recorded per person, case, purpose, timestamp, and channel — CV processing, audio, video, and identity are separate records. Withdrawing one writes a new revision instead of overwriting history.
Missing evidence isn't a zero
When a CV or an answer doesn't evidence a requirement, the result is an explicit review state. Evidence confidence describes how well a claim is supported — not how truthful or suitable a person is.
Retention and data rights
Candidates can request access, correction, consent withdrawal, or deletion, and deletion runs across every case. Raw identity documents and raw interview video are scheduled for deletion by default.
Never inferred, never ranked on
EmotionHonesty as a traitConfidence as personalityMental or psychological stateCultural fitProtected characteristicsGaze-based “cheating”
What gets scored instead is what the candidate actually said against a defined, job-related scenario — with the quote, the skill it maps to, and the limits of the evidence attached. Identity verification never contributes to ranking.
High-volume external hiring
Frontline, field, and branch roles where one posting draws thousands of applicants and a small team has to cover far more CVs than it can read.
Internal mobility
Moving people you already employ, judged on the same evidence rules but reported in readiness language rather than pass or fail.
Hiring that must be defensible
Processes that stay consistent, documented, and explainable months later — to an auditor, to a works council, or to the candidate who was turned down.

See the four assistants
on one of your roles

Send us a position you're hiring for now. We'll configure it with your criteria, your weights, and your interview questions — then walk the whole chain with you, from the sentence you type to the report you'd take into the meeting.

Open AIN Hire
AIN Hire is part of AIN People and shares the orchestration, integrations, memory, and governance layer that runs the rest of the AIN platform. The workspace at ain-hire.anvaya.co.id is live; the large applicant sets inside it are clearly labelled synthetic data, there to show high-volume operations rather than to claim real candidates.