AI and HR: A Match Made in Heaven?

Three reports landed on my LinkedIn this week, all making variations of the same promise: agentic AI is finally going to give HR the seat at the table it's always wanted. HRBPs freed from admin. CHROs orchestrating “human-AI capacity” instead of headcount. Trust, governance and decision rights, all redesigned by HR, for HR. This sounds like a fairytale, but it could also be a pantomime or potentially a horror!

I believe HR earns its seat at the table when the administration and BAU elements of the function work well and stop distracting the business – that's what makes you credible. AI doesn't do that on its own, any more than previous technology did. Implemented well, end to end, it takes the noise away and makes room for us in the strategic conversations. Done badly... well, we all know where that gets us.

Two different kinds of intelligence

Before going further, it's worth being clear about what we're actually comparing. Human intelligence isn't one thing. Psychologists have argued this for decades, and we already accept it in practice – it's exactly why we (I) champion diversity of thinking as one of the most valuable things a team can have. We're living, conscious beings who learn through life, senses, relationships and consequence. We carry values, motivations and judgement, and we're accountable for the choices we make.

Artificial intelligence is something completely different: software that predicts the most probable next output from the patterns in its training data. It can write, reason, recognise patterns and analyse information in ways that look intelligent – but performing well at those tasks doesn't establish an inner experience behind them.

There's a difference between intelligence and consciousness, and agentic AI has given us plenty of the first without any evidence of the second – whatever the scare stories in the media might suggest.

A simple way to put it: humans experience life; AI processes information about it.

That distinction matters here, because HR's traditional strength is built on the human kinds of intelligence. Governing an agentic AI system draws on a different kind altogether – technical, statistical, systems-literate; the ability to read what a model is actually doing, on what data, and why. Being genuinely good at one doesn't automatically transfer to the other. Which is exactly why I don't think the answer to “who should own AI in HR” is simply “HR” – any more than being good with people would qualify someone to audit a payroll integration.

The gap I keep seeing

Throughout my career, I've made a point of giving every team I run – specialist and generalist alike – basic grounding in project management, process design and data literacy and analysis. Not always in depth, sometimes just enough to spot an issue, or know when to question what's in front of them.

I've worked with some brilliant people. But in my experience, most traditional generalist HR professionals don't focus on these skills and aren't wired that way. The ones who do have that instinct for process and data tend to be the ones who move fast – often out of generalist HR altogether, into COE, transformation or commercial roles.

Which is why I'm concerned with these reports. They're selling services and training, so of course the message is that the gap can be closed. But there's a real, structural hole in how much of the profession will ever develop this instinct – not because people aren't capable, but because it's not how most HR careers are built. The people who can already do this work have usually started that journey, or specialised out of generalist HR, years ago. My question is what happens to everyone else.

The numbers say I'm not alone

AIHR's Future-Ready HR Skills research, covering more than 13,000 HR professionals worldwide, ranks digital agility as the lowest-scoring competency in the entire profession – with data literacy not far above it – just think about that for a moment in the context of AI! CIPD's research found that only 6% of UK HR professionals actually use advanced analytical techniques day-to-day. And in Korn Ferry's CHRO survey, 64% of CHROs themselves disagreed that their organisation has strong HR data analytics integrated into business planning. This isn't a training gap you close with a webinar or even a five-day course.

Three reports, three pitches and the same skipped step

Strip away the positioning and each report is selling something slightly different. All three skip the same question.

One – Capgemini Invent's “The Future CHRO in the AI Era,” a global consultancy's point of view aimed at CHROs, says that “no enterprise function is better positioned than HR” to lead this shift. It's a catchy line, and it's selling services on the strength of it. It's also not proven – being close to people and culture isn't the same as understanding how a system decides what to escalate, what to execute, and where accountability sits when it gets something wrong.

Another – AIHR's “HRBP Model Implementation Roadmap,” a step-by-step implementation guide for the HR business partner model – is more honest about the mechanics, which exposes the real problem. The methodology builds the plane before checking who can fly it, only considering this at step four, nine months in. If you know me, you'll know I like that analogy – I've spent enough time around the blades, turbines and engines, and the flight plan, to trust it (and yes, my husband is an aerospace engineer)! But it's the same mistake in a different context: activity, data and the end-to-end process should always come first.

The third – AIHR's “The Post-AI HRBP Model,” also produced for HR professionals worldwide – is the sharpest, and it does note that giving agentic AI to an HR function that hasn't been redesigned “does not liberate HRBPs. It accelerates the status quo.” Its fix is a new job – “AI Governance and Decision Quality Steward” – auditing outputs for bias and scope creep, clarifying accountability, telling a data problem apart from a design flaw apart from a governance gap. That's genuinely technical work, closer to risk and data governance than most HRBPs have even seen, never mind been measured on. It does ask the right closing question – are your people currently capable of this – but it doesn't propose an answer, or a way to find one.

Three different products. One shared assumption: that HR is the right place to build this capability, if only it tries hard enough. But is that setting us up for failure?

The governance challenge and the mess underneath it

I've written before about what happens when we assume a system is “governed” just because the data left it successfully. Naming someone the AI governance owner doesn't make the governance real, any more than declaring a system the single source of truth makes it accurate when it's not integrated or maintained.

Here's the part none of these three reports mention: most of the organisations reading them haven't governed their existing HRIS and ERP data yet, let alone the AI they're thinking they'll pop on top of it. Only 14% of organisations describe their HR technology as coherent and interoperable. If we still can't get “enter once, serve all needs” right for the systems we've had for years, or reconcile who's actually on the payroll between two systems that are both supposed to be telling the truth, just bolting agentic AI on top won't change that. It will make the mess faster, bigger and harder to see.

Jamie Bartlett makes a version of this point in How to Talk to AI, a brilliant book I'd recommend for a lighter read on AI: these systems don't know facts, they predict the most probable next output from whatever they're given. That's not a flaw to be patched – it's how it works. Feed a probability engine clean, governed, reconciled data and it's a genuinely powerful tool. Feed it the same unreconciled HRIS and ERP data most of us have been quietly living with for a decade, and it will produce answers just as confidently. They'll just be wrong in ways that are harder to spot, because they'll sound so plausible.

The GDPR question nobody's asking

There's a risk sitting underneath all of this that none of the three reports mention either – one that gets more attention in Europe than elsewhere. Most of what agentic AI would touch in HR is personal data, often special category data: health, absence, performance, sometimes far more. Feed that into a system where the process and data flows aren't fully understood, and you're not just risking a bad output. You're risking a breach – and it could be a big one.

That's not hypothetical. The ICO is currently consulting on new guidance for automated decision-making and profiling, with recruitment specifically in its sights – its own “Recruitment Rewired” work already found inconsistent human oversight and poor bias monitoring in live use. This is one of the first frontiers AI is being used in HR, and it's taking regulators a while to work through it all – I can remember this starting back in 2018/19. In many European jurisdictions, works councils have rights to be consulted before introducing any system, especially one managing personal data and/or capable of monitoring employee conduct, which most agentic HR tools are, whether that's the intention or not. And how's that conversation with works councils or unions going to go if you can't explain the detail of the data, the process and the usage? Get it wrong, and you don't just have a technology problem – you have an industrial relations one, and it'll be a long while in the resolution.

And here's the part I think gets glossed over: even the best people you have will still be learning how these systems actually behave when they hit reality. That's not a reason to wait. It's a reason to build in a much wider margin for error than feels comfortable – a generous cushion, not a thin one – while everyone involved, however capable, works out where the real risk sits. Get that wrong and the consequences are clear – a breach involving employee data is real, reportable and expensive, in ways a traditional mis-automated process usually isn't.

Understand first. Design second.

I know this is my mantra – that you can't design the job until you've understood the work. The same applies here. You can't design AI governance and decision rights around a workforce, technology ecosystem or data estate you don't understand. And understanding it means an honest capability and data audit, not a competency framework provided by a vendor's training catalogue.

So is it a match made in heaven?

The potential is there. Agentic AI genuinely could free HR from the admin – I've made that case myself, many times. But the potential being there doesn't mean the same people who build our structures around people are automatically the right ones to build them around AI too.

Here's where I am on this, and where I think all three reports get it wrong: I don't think generalist HR should own this. Not because HR people aren't capable of learning it, but because ownership belongs with whoever is already doing the specialised process, data and systems work today, if you have such a team – wherever in the organisation that actually sits. If that's a genuinely capable HR COE or transformation team, fine – hand it to them. If it isn't, the answer isn't to retrain HR into data governance stewards on the side of an already full role. It's to build, or borrow, the specialist capability first, and let HR do what it's actually good at – the human kind of intelligence.

Before we empower anyone to create AI governance and decision design, I think we need an honest answer to a much less polite question than the one these reports ask – not “how do we build this capability,” but “who, today, in most organisations, actually has it – and is it sitting anywhere near the HR org chart?”

My honest answer is: not many, and not usually. And pretending otherwise doesn't help anyone.

So I'm actually worried for the HR function here – it could be pushed into owning something it's not ready for, just because the word “intelligence” happens to cross over from human to artificial.

What are your thoughts? In your organisation, does anyone actually have the depth to own AI governance and decision rights in HR, or not? And have you cleaned up the systems you'd be pointing it at? I'd be really interested in your answer.

Sources

• AIHR – The Future-Ready HR Skills Report
• CIPD – People Analytics: Driving Business Performance with People Data
• Korn Ferry – HR Heal Thyself: CHRO Survey on Gaps in HR Talent
• Jamie Bartlett – How to Talk to AI (And How Not To)
• ICO – draft guidance on automated decision-making and profiling; “Recruitment Rewired” findings
• Capgemini Invent – The Future CHRO in the AI Era (2026)
• AIHR – HRBP Model Implementation Roadmap
• AIHR – The Post-AI HRBP Model