You do not screen 10,000 people one by one — you design a funnel of gates that reduces 10,000 applicants to a number a human can review, and decide deliberately what each gate is allowed to reject. Objective eligibility knockouts remove the largest share first; structured screening at volume narrows further; only a small surviving set reaches a human interview. The gates, not the headcount, are the real design problem.
Why can't you just screen 10,000 people manually?
Manual screening of 10,000 applicants is not a productivity problem to fix with a faster recruiter — it is an arithmetic ceiling. Multiply any realistic minutes-per-screen figure by 10,000 and divide by any realistic team size, and the total working time needed almost always exceeds how long a role can stay open. The calculator below makes that arithmetic visible using your own numbers, not an industry figure.
The calculator's defaults — 8 minutes per screen, 3 recruiters, 6 productive hours a day — are illustrative placeholders only, not benchmarks. This hub publishes no minutes-per-screen benchmark because it has not found one it can verify for the Indian market. Replace every default with your own measured figures before drawing a conclusion.
Screening capacity calculator
At this pace, clearing the queue within 5 working days would need 45 recruiters screening in parallel, not 3.
These defaults are illustrative placeholders, not benchmarks. Replace every field above with your own measured figures before relying on the result.
What are the four gates in a high-volume screening funnel?
This four-gate model separates what should be decided in bulk from what should be decided by a person, in that order.
- 1
Gate 1 — Eligibility knockouts
Remove applicants who are objectively ineligible before any time is spent evaluating fit.
DecidesLocation/commute feasibility, shift and availability match, mandatory documents, legal eligibility to work in the role's jurisdiction.
Must not decideAnything about quality, potential or personality — a knockout gate should never rank the people it keeps.
Design principle, not a measured statistic: this gate should absorb the largest single drop in volume, because it is the cheapest and most defensible to run at full scale.
- 2
Gate 2 — Structured screening at volume
Ask every remaining candidate the same questions in the same order, so the pool that reaches a human is comparable.
DecidesCore role-fit answers — experience, stated availability, basic competency checks — captured consistently.
Must not decideFinal acceptance or rejection on subjective judgement; this gate structures information, it does not make the call.
Design principle: this is the gate most worth automating, because consistency here is what makes every later stage fair to run quickly.
- 3
Gate 3 — Verification and skills evidence
Confirm that what a candidate claimed in gate 2 is actually true before a recruiter's time is spent on it.
DecidesDocument or credential verification, a skills test or work sample where the role justifies one.
Must not decideWhether the candidate is likeable or a 'culture fit' — that judgement belongs later, with a person.
Design principle: keep this gate narrow — verify only what materially changes the hiring decision.
- 4
Gate 4 — Human interview for the survivors
Apply human judgement to the small number of candidates who have already cleared objective checks.
DecidesThe actual hiring decision, including any nuance the earlier gates were deliberately built not to weigh.
Must not decideNothing is off-limits by design — this is the gate built for judgement.
Design principle: if this gate is still reviewing thousands of people, an earlier gate has failed.
Should you use a knockout rule or a ranked score?
The single most consequential design decision in a high-volume funnel is where to use a binary knockout versus a ranked score.
| Dimension | Knockout | Scoring |
|---|---|---|
| How it decides | Pass/fail against an objective rule | Ranks candidates relative to each other |
| Defensibility | Easy to explain and audit | Hard to explain why one candidate scored above another |
| Bias risk | Low, if the rule itself is lawful and objective | Higher — scoring models can encode bias invisibly |
| Cost to run at volume | Very low | Higher — needs calibration, monitoring and review |
This hub's position: use knockouts for anything objective and legally grounded, reserve ranking for building a shortlist among candidates who already passed every knockout, and never rank on an inferred attribute — a guessed trait or a proxy for a protected characteristic — that the employer could not defend if asked.
How do you automate gate 2 without losing fairness?
Gate 2 is usually where volume forces some form of automation. Three broad approaches are in use in India; none is described here with a performance claim, because this hub found no verifiable, India-specific benchmark comparing them.
Async self-serve forms and assessments
Good at: Cheap to run at any volume; candidate completes it on their own time; easy to keep records.
Falls short at: Assumes reliable text literacy and internet access, and cannot read tone, hesitation or context.
Automated voice screening calls
Good at: Works for phone-first, lower-literacy candidate pools and can run in multiple Indian languages.
Falls short at: Language and code-mixed speech coverage varies by vendor and language.
Messaging and chat screening
Good at: Familiar interface for candidates who already use messaging apps; asynchronous and low-pressure.
Falls short at: Harder to verify who is actually answering, and can lose nuance in short text replies.
For more on how automated voice screening works in India, see the Read the multilingual voice AI field guide. To compare vendors directly, see the Compare AI calling vendors.
Mishuk Labs runs automated voice screening calls for exactly this kind of high-volume gate. Mishuk Labs, which publishes this hub, places outbound calls to candidates and runs structured automated voice screening in 10+ Indian languages, returning a scored shortlist to the recruiting team. That fits gate 2 directly: when a single role draws thousands of applicants, every one of them needs the same questions asked the same way, which is the consistency an automated calling flow is built to hold at volume. See the full vendor comparison.
Readers comparing options for gate 2 can see how form-based, chat-based and voice-based approaches differ in this hub's dedicated vendor comparison.
Mishuk Labs publishes this hub. See Mishuk Labs for product details.
What decisions should stay human?
Some decisions should never be pushed onto an automated gate purely because of volume. This hub treats the following as non-negotiable:
- Final rejection of any borderline or close-call candidate.
- Any situation involving a disability or a request for reasonable accommodation.
- Any decision that depends on context an automated gate cannot see — a resume gap, an unusual answer, a circumstance the candidate explains.
- Any decision the employer could not explain in plain language if the candidate asked why.
What does India actually regulate here?
Screening 10,000 applicants touches two separate questions: whether automated hiring decisions are specifically regulated in India, and what obligations apply simply from holding that many candidates' data. This is an editorial summary, not legal advice.
No Article 22 equivalent in India's DPDP Act
India's Digital Personal Data Protection Act, 2023 contains no equivalent to Article 22 of the EU GDPR: no general right against a solely automated decision, no right to human intervention or review, no right to an explanation, and no dedicated regulation of profiling.
What to do: Do not assume a GDPR-style automated-decision right applies in India — it does not currently exist in statute.
No India-specific AI-hiring statute (as of 2026)
As of 2026, India has no generally applicable statute, regulation or binding government guideline specifically governing AI or automated tools used in hiring, résumé ranking or candidate selection.
What to do: Do not rely on an AI-hiring-specific regulation to define what is permitted — none exists yet; general law still applies.
General non-discrimination law still applies
General protections still apply regardless of automation: the Rights of Persons with Disabilities Act, 2016; the Transgender Persons (Protection of Rights) Act, 2019; and, for public employment, Articles 15–16 of the Constitution of India.
What to do: Screen every automated gate's rules against these protections before deploying it at volume, not after a complaint.
IndiaAI/MeitY responsible-AI material is guidance, not law
MeitY/IndiaAI responsible-AI materials discuss fairness, non-discrimination, human supervision and grievance redress in AI hiring, but these are guidance and discussion documents — not enforceable hiring regulation.
What to do: Treat this material as good practice to adopt voluntarily, not as a compliance requirement you can point to.
DPDP obligations that do apply to a 10,000-candidate dataset
Holding 10,000 candidates' data still triggers standard DPDP obligations: clear notice, a specific stated purpose, purpose limitation, and retention tied to that purpose with deletion once it is served. The DPDP Rules, 2025 were notified 13 November 2025, with the bulk of substantive notice, consent, security, retention and rights duties phasing in around 13–14 May 2027.
What to do: Set a retention period for the applicant dataset now, tied to the hiring cycle's actual purpose, rather than waiting for the 2027 date.
The compliance gap is not a licence to skip explanation (hub analysis)
This is the hub's own analysis, not a stated legal position. The absence of an Article 22-style right in India is a compliance gap, not a licence: an employer that cannot explain why a specific candidate was rejected still carries reputational risk and exposure under general non-discrimination law, even without an AI-specific statute to point to.
What to do: Keep a decision log recording which gate rejected each candidate and why, and take legal counsel before relying on any interpretation here — this is not legal advice.
This is an editorial compliance summary, not legal advice, current as of 2026. Laws, rules and commencement dates change — verify every point against the current statute and take qualified counsel before setting policy.
Frequently asked questions
Sources & methodology
Base statute; contains no Article 22-style automated-decision right.
indiacode.nic.inPhased commencement of notice/consent/retention duties, ~13–14 May 2027.
meity.gov.in · Notified 13 November 2025General disability non-discrimination protection, applies regardless of automation.
indiacode.nic.inGeneral employment non-discrimination protection.
indiacode.nic.inArticles 15–16, equality and non-discrimination for public employment.
indiacode.nic.inGuidance on fairness, human supervision and grievance redress; not enforceable regulation.
indiaai.gov.inThis article cites no productivity, conversion or cost benchmark for screening, because this hub found none it could verify for the Indian market; the capacity calculator's default values are illustrative placeholders, not benchmarks, designed to be replaced with the reader's own measured figures. The four-gate funnel model and the knockout-versus-scoring position are this hub's own analysis and framework, not a third-party standard. The legal position in the compliance section reflects statute and published guidance as of 2026, is not legal advice, and should be reconfirmed against current law and with qualified counsel before setting policy.