This article discusses building a fair AI interview process from a talent acquisition perspective. It references the UK regulatory environment as context, but is not legal advice. For specific compliance decisions, consult a qualified UK data protection or employment law specialist.
Building a fair AI interview process is quickly becoming the difference between UK SMEs who keep their strongest candidates and the ones who lose them without ever knowing why.
A founder rang me last month, mid-hiring-round, and the first sentence was, “Sabiha, three of our shortlist just pulled out. I don’t know what happened.” What had happened, once we walked through it, was that the third stage of her process involved a recorded AI interview. Nobody had told the candidates that in advance. Two of them worked it out from the platform’s branding and one of them realised only when the video started.
She was not doing anything unusual. According to the Greenhouse 2026 Candidate AI Interview Report, 82% of UK candidates who have been through an AI interview were never clearly told upfront that AI would be evaluating them, and 24% only found out once the interview had begun. Thirty per cent of UK candidates have walked away from a hiring process specifically because it involved an AI interview. Another 19% say they would.
That is not an AI problem. That is a design problem. And it is a fixable one.
The interview does not become AI or human. It becomes AI-assisted. That distinction is the whole thing.
Most of the UK SMEs I speak with think of AI interviewing as a binary. Either you bring in a video-AI platform that scores candidates for you, or you keep everything human and old-school. Neither of those is what a good process actually looks like in 2026.
A fair AI-assisted interview keeps the human doing the judging. AI handles the parts humans are bad at, like scheduling, transcription, note-taking, consistency of question order, and post-interview summary. The candidate still meets a person. The decision still sits with a person. What changes is the scaffolding around it.
This matters because the research on interview quality has not shifted. Sackett’s 2022 meta-analysis reconfirmed what earlier work had suggested: structured interviews remain the strongest single predictor of job performance. The evidence has not changed. AI is a way of making the structure easier to hold, not a replacement for it.
Here is the framework I take SMEs through when they ask me how to introduce AI into their interview stage without losing candidates or ending up on the wrong side of the ICO.
Before you configure anything, write it down. Which parts of the interview process is AI touching, and what is it doing there?
The safest first uses are administrative. Interview scheduling. Transcription. Structured note capture during the conversation. Post-interview summary against your rubric. These are low-risk because they support the interviewer rather than replace them.
The higher-risk uses are the ones that need the most care. Scoring candidate responses. Ranking. Auto-progressing or auto-rejecting. If you are using AI for these, you are in the territory the ICO’s Recruitment Rewired report flagged in March 2026. Its central finding was that many UK employers believed they were using AI as decision support, when in practice the tools were making the decisions and the human was rubber-stamping the output.
Write the list. Share it with your hiring managers. If a tool is doing something that is not on the list, take it off.
Not at the point they click record. Not in a privacy notice buried three clicks deep. In the invitation email that offers them the interview slot.
The disclosure needs three things. What AI is doing at this stage. What it is measuring. Who makes the final decision. Three sentences will do it.
This is the single change that closes the biggest trust gap in the Greenhouse data. Only 1 in 10 UK candidates say their employer had clear AI policies, and 59% think disclosure should be a legal requirement. The direction of travel is clear, and the ICO is signalling firmly that transparency is not optional. Getting ahead of that costs nothing.
A structured interview means the same questions, in the same order, evaluated against the same rubric, for every candidate for the role. This is where AI actually earns its place.
Use it to prompt interviewers with the next question. Use it to transcribe answers so the interviewer can stay present and listen properly instead of scribbling. Use it to generate a first-pass summary against your rubric after the call, which the interviewer then reviews and adjusts.
What you avoid is the AI producing a score the interviewer never questions. The moment the human treats the AI output as the answer rather than a draft, the human involvement has stopped being meaningful in any real sense. Under the reformed Articles 22A–22D of the UK GDPR, in force since 5 February 2026, meaningful human involvement is the operative test for whether a decision counts as solely automated. The ICO’s own guidance signals that a hiring manager glancing at a shortlist they cannot really override is exactly the pattern it intends to scrutinise. A qualified UK data protection specialist can advise on where a specific process sits against that test.
Meaningful human involvement is not a checkbox. It means someone who can read the AI output, understand how it was produced, disagree with it, and change the outcome, with the authority and the time to actually do so.
Practically, for an SME, that looks like a named reviewer for every stage that involves AI. They have seen the training on how the tool works. They can articulate what the score reflects and what it does not. They have a workload that lets them spend real time on each candidate rather than three minutes and a nod.
If the honest answer is that your hiring manager is too stretched to review 40 AI-shortlisted candidates properly, the fix is not a better AI. The fix is a shorter shortlist, or a different stage design.
Two things need to exist by the time you launch the process. A named point of contact who can answer questions about how AI was used in a decision. And, where practical, the option for a candidate to request a human-led interview instead.
The second one worries SMEs the most. Almost always, when I actually work through it with them, hardly any candidates request the human route. What matters is that it exists. The Greenhouse data found that 45% of UK candidates want the option to request a human interview. Offering it, even when few take it, does more for candidate trust than any brand campaign.
Run this as a fair AI interview process end to end, and three things change.
Candidates stop walking out of your process. The 30% withdrawal figure Greenhouse recorded was driven overwhelmingly by ambush and opacity, not by AI itself. Only 19% of UK candidates say they want less AI in hiring. The rest want the same amount or more, provided they know it is there and how it works.
Your interview data gets better. Structured, transcribed, rubric-scored interviews with a real human decision on top produce a defensible audit trail that a spreadsheet of gut-feel notes never can. If you are ever asked to demonstrate fairness in your process, you have something to show.
And you stay on the right side of the regulatory direction of travel. The ICO wrote to 16 named organisations after Recruitment Rewired. All 16 committed to act. The final statutory Code of Practice on AI and automated decision-making is expected in the coming months. Building the framework now costs less than retrofitting it after the code lands.
Pick one open role. Write your one-page disclosure statement. Rewrite the interview invitation email so that disclosure sits above the fold. Choose one AI function to add and one to remove. Run the next interview through it, then debrief the interviewer honestly on whether the human involvement was meaningful.
It is not glamorous work. It is quiet, procedural, structural work. It is also, in my experience across 16 years and four countries of hiring, the difference between a fair AI interview process that hires well and one that quietly loses your best candidates before they ever meet you.
This is a data protection question, not a talent acquisition one, so Sabiha does not offer legal advice on it. What can be said as context: since 5 February 2026, Articles 22A–22D of the UK GDPR govern automated decision-making, replacing the old Article 22. The ICO’s Recruitment Rewired findings and the direction of travel in its draft guidance both signal that genuine, meaningful human involvement is the factor regulators are focused on. For a view on where a specific interview process sits against that framework, a qualified UK data protection or employment law specialist can advise.
Transparency is a core expectation under UK data protection law and a clear expectation from candidates. In the 2026 Greenhouse survey, 59% of UK candidates said AI disclosure should be a legal requirement, and the ICO’s Recruitment Rewired findings pointed to transparency as one of the most common gaps in current practice. The safest and most trust-building approach is to disclose in the interview invitation, before the candidate has to decide whether to attend.
AI-assisted interviewing means AI supports a human interviewer with scheduling, transcription, note-taking, structured question prompts, or first-pass summaries. The human makes the assessment. AI-led interviewing means the AI conducts the interview and, in many cases, scores or ranks candidates with limited human review. The two carry very different regulatory exposure and very different candidate trust profiles.
In the Greenhouse 2026 Candidate AI Interview Report of nearly 3,000 active jobseekers, 30% of UK candidates said they had already walked away from a hiring process because it involved an AI interview, and a further 19% said they would. The report attributes this less to AI itself than to how it is being introduced without disclosure or context.
Begin with administrative support functions rather than candidate assessment. Interview scheduling, live transcription, note capture and post-interview summary against a rubric are lower-risk starting points that build genuine familiarity with the tools before you introduce anything that touches scoring or ranking. This staged approach is the fastest route to a genuinely fair AI interview process.
Sabiha is a Talent Acquisition Director, Speaker and Author with 16+ years of international hiring experience across the UK, Dubai, South Africa and Malaysia. She has advised 300+ UK businesses on hiring and retention and was shortlisted for Best Career Coach UK by the Career Development Institute. Her book, How to Use AI to Win Talent and Retain People (Trotman, CIPD-aligned), is due in Autumn 2026. Sabiha writes and speaks on AI in hiring, workforce strategy and retention design for UK SMEs at meetsabiha.com.
Global Talent. Ethical AI. Strategic Hiring. Sustainable Retention.
