FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Recruitment agencies
An agency that offers AI-scored video interviews is responsible for how candidates are assessed, including reasonable adjustments and automated decision safeguards. Clients buy these services for speed, but the agency can carry the risk if scoring is unfair.
Part of: AI in recruitment agencies
In short
Some agencies offer clients one-way video interviews scored by AI, which rate answers, speech or behaviour. The ICO's 2024 audit and its 2026 Recruitment rewired report raise concerns about profiling and solely automated hiring decisions. An agency is an employment service-provider under the Equality Act and must make reasonable adjustments, which is hard if a tool penalises a speech impairment or atypical eye contact. A score that decides who progresses without meaningful human review triggers Article 22C safeguards. Discrimination claims may fall outside PI, and the client's EPL may not cover the agency.
Candidates record answers to set questions, and the platform transcribes and scores them. Some tools rate content only; others analyse tone, pace, facial expression or eye contact. Agencies may sell this as a screening service, often for graduate or volume roles, and pass a ranked list to the client. The agency may not know exactly what the tool measures, because the model is the supplier's.
The Manjang v Uber Eats case, where automated facial recognition checks led to a race discrimination claim that settled after the tribunal refused to strike it out, shows how facial analysis tools can be challenged.
Illustrative: an agency runs AI-scored video interviews for a client's graduate scheme. A candidate with a speech impairment scores in the bottom band and is rejected automatically. She was not told how scoring worked or offered an alternative format. She brings a disability discrimination claim against the agency and the client and asks the ICO to look at the automated rejection. The client points to the agency's service agreement, which said the tool was fair and validated.
When an agency resells a scoring platform, it often becomes the party the candidate deals with and the party the client relies on. Service descriptions written by sales teams can overstate what the tool does, for example by calling it validated, objective or compliant. Those statements may become contractual promises. Before offering the service, the agency should know who is controller for the video and score data, what the supplier will stand behind if a claim arises, and whether its own terms with the client limit liability for the tool's output. APSCo Global's 2025 guide recommends risk assessing AI tools and holding suppliers accountable for security.
Employment practices liability insurance usually protects a business against claims by its own employees and, under some wordings, job applicants to that business. The client's EPL, if any, is written for the client, not the agency. The agency's PI may exclude or limit discrimination claims. A separate claim by the client against the agency, for example that the scoring service was negligently designed or misdescribed, may fall within PI, subject to the wording and any contractual warranty. This gap is worth closing before you sell the service. See EPL explained and AI bias and discrimination claims.
Apex arranges PI for recruitment agencies; a broker can review how your PI treats assessment services and discrimination allegations. See the AI in recruitment agencies hub, AI CV screening and telling candidates you use AI.
They are not banned, but the agency must meet UK GDPR, including Article 22C safeguards where a score decides the outcome without meaningful human involvement, and the Equality Act, including reasonable adjustments. The ICO has raised concerns about automated hiring decisions and inference of characteristics.
Where a candidate's disability puts them at a disadvantage, the agency as an employment service-provider has a duty to make reasonable adjustments. That may mean a different format or human assessment. Tell candidates in advance how to ask.
Usually not. An EPL policy is written for the policyholder and its own employees, and sometimes its applicants. The agency should check its own PI, which may exclude or limit discrimination, and its contract with the client.
Where the score decides who progresses, meaningful human review reduces the risk that the decision is solely automated and gives a route for Article 22C challenges. The reviewer should see the recording and be able to change the outcome, not just approve the ranking.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.
