How are UK businesses using AI in HR?

HR has the broadest AI deployment of any business function. It also has the weakest proof it’s working. In aibl’s survey of 755 UK mid-market business leaders, including 152 HR, people and L&D leaders, HR runs AI almost everywhere, yet a written policy nobody enforces returns just 16% measurable ROI, below the 33% reported with no policy at all.

HR deploys AI more widely than any function, and proves it least

HR has AI running in more places than any other function in the survey. Screening, scheduling, drafting, onboarding, internal queries. The tools are in.

The return is the harder question. Across the whole sample, about half of the UK mid-market businesses surveyed can show a measurable return on AI, and HR sits inside that same proof gap rather than ahead of it. Wide deployment hasn’t converted into wide, provable value.

The reason isn’t the tools. It’s whether anyone owns the outcome and checks the number. In HR that discipline is rarer than the software, which is how a function can be busy with AI and still unable to say what it returned.

The policy trap is sharpest in HR

A written AI policy that nobody follows can be worse than having none, and nowhere is that clearer than HR. In aibl’s survey of 755 UK mid-market business leaders, HR functions with a policy applied inconsistently report 16% measurable ROI. HR functions with no policy at all report 33%.

Line up the five governance levels for HR and the dip is stark: 33% with no governance, 23% informal, 16% defined but inconsistent, 56% formal, 82% mature and embedded. The middle rung scores lower than the bottom one.

The cause is false confidence. A framework on the intranet feels like the job is done, so the money to enforce it never gets spent. The fix is enforcement, not a better document. Moving HR from a policy-on-paper to an enforced one is the single biggest governance gain in the whole survey, worth 40 points. Our piece on why an AI policy nobody follows is worse than none covers the pattern in full.

Who should own AI in HR? Not the CEO

HR is the one function where the CEO shouldn’t own AI. In aibl’s survey of 755 UK mid-market business leaders, Central IT ownership of HR AI returns 59% measurable ROI, ahead of CEO ownership at 45%.

Everywhere else, ownership at the top wins. HR is the exception because its AI depends on technical controls IT already runs: access to employee records, recruitment data, and the audit trail behind an automated decision. Executive sponsorship alone doesn’t provide those.

So the model that works in HR is a pair. A named owner accountable for the outcome, and Central IT running the controls underneath. Our guide to who should own AI sets out the wider ownership data.

AI in recruitment carries the highest compliance exposure

AI in hiring is where the reputational and legal risk concentrates, and where the policy gap becomes a compliance gap. The ICO’s November 2024 audit of AI tools used in recruitment made 296 recommendations.

Its central finding matters for every CHRO: passively accepting an AI-generated shortlist, without a real check, does not count as meaningful human oversight. A human name on the process isn’t enough if no human actually reviewed the decision.

For HR, that turns the gap between policy and practice into more than a return problem. A documented human review of AI hiring decisions is the compliance defence. Make it a recorded step, not an assumption.

What this means for the CHRO

Start by pairing HR with Central IT on the controls, so the technical guardrails behind employee data and automated decisions are owned properly. That single move addresses the ownership finding and the compliance one at the same time.

Then enforce the policy you already have rather than writing a new one. Name who owns AI outcomes in HR, and make human review of any AI hiring decision a documented step. That’s the 40-point move.

And build capability in your own team. Across the survey, training-led firms report 73% measurable ROI against 14% for those who leaned on consultants (a 150-person sub-sample). Our post on AI training for business covers how to build it.

Frequently asked questions

Read the full research

This is the HR cut of State of UK AI Adoption 2026, aibl’s benchmark of 755 UK mid-market business leaders, in partnership with Executive Summary. The full report covers every function, the governance ladder behind these numbers, and the four moves that separate the firms proving a return from the rest.

Read the full State of UK AI Adoption 2026 report →

If you are accountable for AI in HR without a team to build it, the AI Enablement Directory lists UK partners by category, including AI training and capability building, and AI governance and risk. New findings land each Friday in aiblBRIEF, our free weekly newsletter.

About the research: the State of UK AI Adoption Survey 2026 was run with aibl’s research partner Executive Summary. It covers 755 UK mid-market business leaders in organisations with revenues of £20m to £500m, fielded January to March 2026. Every figure above describes the leaders surveyed.