AI in hiring: what HR leaders need to know about governance and trust

Most HR leaders using AI in hiring will tell you they’re doing it responsibly. Fewer can explain exactly what that means in practice – who owns the decisions, how consistency is enforced, what candidates are told and why. That gap between confident adoption and actual governance is what Assessio’s Maturity Gap research set out to measure.

This spring, we put the findings to two people living the reality from very different vantage points.

Kasper Abrahamsen, Global Recruitment Manager at IO Interactive, joined session one to talk about what responsible implementation looks like from inside a recruitment team.

In session two, Katrina Collier, author of The Robot-proof Recruiter and Reboot Hiring, argued that the human cost of getting this wrong is already taking effect. Both conversations were uncomfortable in the right ways. Here’s what stayed with us.

Start questioning how you are using AI

Most organisations haven’t made this shift yet. What Kasper Abrahamsen describes as a “drop in, drop out” reality will be recognisable to most recruitment teams: individuals using different tools, different prompts, producing inconsistent outputs, with no one accountable for the aggregate result. AI is doing work, but the work isn’t governed.

“We need to stop talking about ‘are we using AI’ and start talking about how do we use it – and then how do we use it consistently and responsibly, and then how do we scale it.” – Kasper Abrahamsen

At IO Interactive, the response was to work backwards from the process rather than forwards from the technology. What decision needs to be made at each stage of recruitment? What is each stage actually trying to assess? Only once those questions have real answers does it make sense to ask how AI can support them, always as a source of insight, never as the one making the call.

The application flood is already here

Over Christmas, Kasper’s colleague Lucas posted two design roles and returned in January to 1,300 applications. All of them polished. All of them effectively identical. A team of four recruiters, one on maternity leave, faced an impossible task.

“We’ve been teaching people for years to tailor their CV and cover letter to the job description. They’re doing that now using AI.” — Kasper Abrahamsen

The consequence is a process under real strain. AI-written applications are being reviewed by AI screening systems, with no meaningful signal in the middle. Volume has broken the model that most recruitment processes were built around, and few organisations have a clear answer to what replaces it.

AI amplifies bad briefings

Most hiring processes fail before a single candidate has applied. That’s Katrina Collier’s argument, and she traces it to a specific point of failure. The intake session between recruiter and hiring manager is consistently the part of the process that gets the least attention.

A manager whose team loses someone will often panic, reach for an old job description, or generate a new one with ChatGPT. The recruiter goes to market with a starting point that’s already wrong. Add AI to that process and you don’t fix the problem, you just scale it.

“You need to know what you’re looking for before you go out and choose your tools to find someone.” — Katrina Collier

“But humans are biased too” is not an answer

AI tools that promise to remove human bias are trained on human-generated data, carrying the implicit biases of whoever created it and whoever prompted it. When Katrina presses tool providers on how they’re mitigating against this, the response she often gets is: “But humans are biased.” Her reaction: “Exactly — that’s why I’m asking.”

The concern is specific as well as general. There are documented examples of large language models appearing to favour applications written by the same model doing the screening — a ChatGPT-written CV scoring better with a ChatGPT-powered screener.

Both guests arrive at the same principle independently: AI should suggest, not decide. Humans remain responsible for consequential calls. That is the design requirement, not an interim position.

👉 For a deeper look at how bias enters AI hiring systems and what organisations can do about it, see Assessio’s guide to AI bias in hiring

86% of candidates who don’t hear back become down or depressed

That figure comes from Tribepad’s End Ghosting campaign, and Katrina used it to put a number on what automated, unaccountable hiring processes cost the people on the receiving end. Recruitment teams have had the tools to close every application for years. The problem has never been capability — it’s that many recruiters haven’t been trained to handle hostile responses to rejections, so they avoid giving them at all.

Her prescription is practical: give everyone closure, even when the message is brief. When delivering personal feedback, use the word “demonstrate,” borrowed from leadership coach Sue Ingraham. Don’t tell a candidate they lack a skill, tell them they didn’t demonstrate it in the interview. It’s harder to dispute, more useful to the candidate, and less likely to provoke the defensiveness that makes recruiters reluctant to give feedback in the first place.

Regarding automated rejections, it’s better to own them. Acknowledge they’re cold, explain why they exist, and use the space to give candidates something they can act on. The goal is to be recognisably human, even in a template.

The path forward is clearer than it looks

Both sessions surfaced real problems — ungoverned tools, broken intake processes, candidates left without closure. But neither Kasper nor Katrina was pessimistic. The disruption, as Kasper put it, is forcing a complete rethink of recruitment methods that in many cases haven’t been meaningfully challenged in decades. That’s uncomfortable, but it’s also an opportunity.

The organisations that will get this right aren’t the ones with the most sophisticated AI tools. They’re the ones that fix the process first, govern the technology deliberately, and are honest with candidates about how and where automation plays a role. Transparency and accountability aren’t obstacles to AI adoption in hiring — they’re what makes it truly sustainable.

If you want to go deeper on what HR leaders across Europe are currently doing with AI, where the gaps are widest, and what the path to responsible implementation looks like in practice, our full white paper is available now.

👉 Download the Maturity Gap: The HR leader’s guide to AI adoption, governance and trust