ATPAVENIR TECHNOLOGY PARTNERS
All work

03Recruitment · Policy and change

Faster to shortlist. Fair by design.

A national recruitment group wanted AI in its hiring process without losing the judgement of its consultants or the trust of its candidates.

Client
A national recruitment group in the UK
Scale
Many desks and offices across the UK
Duration
Five months, including an eight-week pilot
Team
Three from ATP, with an employment law specialist from our network

A representative example of our work. Client details are withheld.

A quiet interview room at dusk, two chairs at a small table beside a glass wall

The problem

What it was costing.

Consultants spent most of their day reading CVs and chasing diaries, and good candidates went to faster competitors. Clients were asking whether AI was in the process; the honest answer was that some consultants used it and no one had tested it for bias.

The solution

AI that assists consultants, tested for fairness.

We set the group’s policy on AI in hiring, redesigned the journey from application to shortlist, and brought its consultants with us from the first workshop.

Policy

  • People decide.AI may summarise and suggest; a consultant makes every shortlist and every rejection.
  • Fairness testing.Every tool tested for adverse impact across protected characteristics before use, and every quarter after.
  • Open with candidates.Candidates told where AI is used in their application and how to ask for a human review.
  • Data protection.Lawful basis, retention and the rules on automated decisions under UK GDPR, set out for each step.
  • Supplier standards.Due diligence for every AI recruitment tool, drawn from the government’s guidance on responsible AI in recruitment.

Process

  • The journey mapped.Every step, handover and wait from application to shortlist, timed on real vacancies.
  • CV summaries.Skills and experience drawn out against the role’s criteria, with the evidence quoted from the CV.
  • Job adverts.Drafted from the role’s requirements and checked for language that puts good candidates off.
  • Interview scheduling.Candidates book from live calendars, with reminders and rescheduling handled for them.
  • Client reporting.Weekly shortlist reports compiled from the system, not assembled by hand.

Change

  • Consultants in the design.Working sessions with consultants from every desk shaped each tool before it was built.
  • A pilot on two desks.Eight weeks, measured against matched desks working the old way.
  • Training before access.A short course on using AI well, and on its limits, completed before anyone is given the tools.
  • Candidates heard.Feedback at each stage, read every month alongside the fairness results.
  • Handover.An internal governance group owns the policy, the fairness tests and the supplier reviews.

The payoff

Measured against the baseline.

Not everything moved. Candidate satisfaction held steady rather than rising. The group’s next step is the feedback a candidate receives after a rejection.

What’s next

Built to carry on without us.

The group is applying the same tests to sourcing and to its client-facing market reports, and publishes a short statement for candidates on how AI is used in its hiring.

Talk to usEvery first conversation is with a partner, and costs nothing.

Ask us about AI in your hiring