ReviewDesk guide · By the Ingenious AI team · Published 14 November 2026 · Reviewed 14 November 2026
What are a staffing agency's duties when it screens candidates with automated tools?
Generate a free adverse-outcome letter — no sign-up Build a free pre-use notice See pricing
The short answer. The candidate faces one process, but two organisations stand behind it — so the duties must be allocated in writing before the first placement: whose notice the candidate sees, whose name the explanation carries, who staffs the review, and how the records divide when the engagement ends. Run per-client workflows on a Team desk so each employer's file stays separate, and let the agency's bench operate all of them to one standard.
What is staffing-agency screening duties?
Staffing-agency ADMT duties are the automated-decision obligations arising when an agency screens, ranks or assesses candidates for client employers: pre-use notice to candidates, adverse-outcome explanations, correction and human-review routes, and per-client records. The duties attach to the decisions about candidates; agency contracts allocate the machinery, while the candidate experiences — and judges — a single process.
Why does agency screening multiply the duty problem?
Because every element doubles. The tool may be the agency's, the criteria the client's, the decision shared in proportions nobody documented. The candidate's notice must describe that reality honestly — who is screening, for whom, with what say in the outcome — or it describes a fiction that the first review request will expose. And volume multiplies too: an agency screening for thirty employers runs more consequential decisions in a week than most employers run in a year, across several states' regimes at once.
The multiplication is manageable exactly because it is systematic. The pattern per client is the one from the hiring guide; the agency's craft is running that pattern thirty times without cross-contamination — per-client workflows, per-client records, one bench discipline — the Team-desk model of the agency guide applied to candidate screening.
What must the agency-client contract allocate?
Five things, in writing, before launch. The notice: whose name and branding, placed in which journey, drafted from whose criteria description. The screening configuration: factors and thresholds per client, with change control — a client who tightens criteria mid-campaign changes the notice's truth. The adverse letters: generated per client from the screening record, on the client's workflow, with the client's review offer. The review bench: who reviews, with whose authority to advance a candidate into the client's process. And the records: per-client files, exportable to the client, with the exit mechanics defined while everyone is friendly.
Silence on any of the five becomes a dispute with a candidate's deadline attached. Agencies that have lived one such dispute write the allocation schedule into their standard terms; agencies that haven't are invited to learn from this paragraph instead, at a considerable discount.
How does the candidate's notice work across clients?
Per engagement, in the client's colours where the candidate's relationship is with the client, and with the agency's role stated plainly either way: 'ClearMark screens applicants on behalf of Dovecote' is one honest sentence that prevents a dozen confused requests. Where the agency hires for its own pool — candidates engaged by the agency and assigned out — the agency is the employer for notice purposes and its own workflows carry the duties. The two shapes must never blur inside one candidate journey; a candidate told they are applying to the agency who is actually being screened for a client has been misdescribed to, whatever the paperwork says.
Jurisdiction follows the candidate and the work, per the routing principles in the jurisdictions guide. Agencies recruiting nationally should expect all three regimes in play simultaneously and build the routing into intake, rather than discovering a California candidate inside an Illinois workflow at review time.
How do reviews work when the client makes the final call?
In two tiers, honestly described. The agency's review re-examines the screening — the record, the factors, the parsing, the criteria as configured — and can restore a wrongly screened-out candidate to the client's process. The client's own decision thereafter (interview, offer) is the client's, with its own duties where the client uses its own tools. The candidate's letter should say which tier their request reaches and what it can change, because a review that can restore screening but not promise an interview must not be sold as either more or less than it is.
Authority is the design constraint: agency reviewers need the client's advance grant to restore candidates into the process, or every successful review becomes a negotiation. Write the grant into the allocation schedule — restoration on screening error is automatic; restoration on judgement calls follows a named client contact's confirmation within a set time. Candidates experience one process; the tiers are choreography behind it, and choreography is rehearsed in contracts, not improvised per case.
How do agency records stay per-client under one bench?
By the platform's business boundary, worked without exception: each client's cases, letters and traces live in that client's workspace; reviewers hold per-client access; the aggregated board shows the bench its clocks without mixing files. Sampling and calibration happen per client — a reviewer excellent on logistics screening is not thereby calibrated on care roles — and the monthly reporting to each client is computed from their cases alone, per the discipline in the history guide.
The agency's cross-client asset is pattern, anonymised: correction sources, factor behaviours, vendor data quality — the intelligence that makes the agency better at screening than any single client could be. That asset is legitimate precisely because case content never crosses. Guard the line in tooling and in habit; the first time a reviewer pastes one client's case into another's channel, the line has already moved.
What does great agency practice look like commercially?
It looks like the screening layer clients cannot build themselves: notices live per client, letters a candidate can act on, reviews answered in days, a quarterly evidence conversation per client that starts from their own pack. Agencies compete historically on candidate supply and speed; the regimes hand them a third axis — demonstrable fairness machinery — that clients' counsel increasingly score in vendor selection. The agency that brings a live desk to the pitch, with coverage counted from real cases under the coverage discipline, is selling the one thing in the meeting that cannot be counterfeited overnight.
Build it on the first client properly, as the reseller guide counsels, and let the pack close the second. The bench, the rhythm and the records compound across clients in a way brochures never do — because in this market, unusually, the operational truth is the marketing.
Your five actions for staffing-agency screening duties
- Allocate the five duties in the client contract before launch: notice, configuration, letters, review authority, records and exit.
- Give every candidate journey one honest description of who screens, for whom, with what say in the outcome.
- Build two-tier review with the restoration grant written down — screening restored automatically on error.
- Hold all records per client workspace; let only anonymised patterns cross between clients.
- Report and sell from live evidence: per-client packs, coverage from real cases, quarterly pattern conversations.
Worked example: an agency restores a candidate its own parser lost
A fictional agency — Summit Staffing, from the evidence guide — screens warehouse candidates for a retail client. A candidate's application is screened out: the parser reads his forklift certification as expired, missing the renewal page. His review request, filed from the client-branded notice page, reaches the agency bench with the case assembled. The reviewer spots the second page in the original upload, corrects the record, and the restoration grant does the rest: the candidate re-enters the client's process the same day, interviews that week, and is hired.
The client's monthly pack tells the story in their numbers alone: 212 screened, 9 requests, 2 restorations, median answer three days. Summit's quarterly pattern review adds the anonymised layer: parser renewal-page misses account for a third of corrections across its logistics clients, and the intake form gains a renewal-date field. One candidate's restored job, thirty clients' cleaner intake — the agency model working exactly as its structure promises, because the allocation was written before it was needed.
Frequently asked questions
Who owes the candidate notice when an agency screens?
The candidate must be told honestly who screens, for whom, and with what say in the outcome — allocated in the agency-client contract before launch. Where the agency employs the candidate itself, the agency carries the employer duties directly.
Can an agency reviewer restore a screened-out candidate?
Yes, if the client's allocation schedule grants it — restoration on screening error should be automatic, with judgement-call restorations confirmed by a named client contact inside a set time. Without the grant, every successful review becomes a negotiation.
How are records kept when one bench serves many clients?
Per client workspace, without exception: cases, letters and traces stay in the client's business, reviewers hold per-client access, and only anonymised patterns — never case content — inform the agency's cross-client learning.
Which plan fits an agency?
Team: multiple businesses under one desk, everything in Pro, with licence terms aimed at agencies and resellers. Each client's workflows, notices and records stay separate inside it.
Is this guide legal advice?
No. ReviewDesk is templates and workflow tooling, not legal advice. The allocation model is ReviewDesk's operating pattern; consult counsel for your client contracts.
Keep reading: How do automated-decision duties apply to gig platforms and marketplaces? · Why does ReviewDesk render missing evidence as 'not recorded' — and never invent it? · How do you integrate decision traces through the ReviewDesk workspace API? · Who should own an ADMT programme — owner, reviewers and record-keepers? · What are the key ADMT deadlines — Illinois now, Colorado and California 2027, the EU in December 2027? · All ReviewDesk guides · ReviewDesk home · Try the free desk · Free letter generator · Free notice builder · Pricing
ReviewDesk is built by Ingenious AI. Templates and workflow tools — not legal advice.