HR has a big image problem.

Last week myself and Vanessa Sanchez from Nova spent three days at Unleash in Paris. If you work in HR and you’ve never been to Unleash (or any HR Tech Conference) it can be a fairly daunting and immersive experience as you’re trying to get your head around what all the different vendors do and which speakers are worth your attention.

I’ve read a lot of really interesting takeaway posts from the likes of Riina Hellström, Sharon O’Dea and Oliver Kasper about the impact AI is going to have on HR data, what the future of work will look like and where the market is heading, but one thing I haven’t really seen or heard too much of is how companies are going to have to start fundamentally rethinking how their People Operations and HR Teams are going to be designed to ensure they’re ready for this new all conquering AI future. It seems HR is now being asked to do three jobs at once. Or is about to be. And this is on top of a calendar that is already stacked.

1. Run the people agenda. No change there.
2. New: build and direct the AI agents that are being designed to do more of the admin work.
3. New: act as a data controller of the most sensitive data in the business, just as that data starts flowing into places no HRIS is designed for.

Have companies started thinking about 2 and 3? I’m not sure HR is attracting the right candidates for the second and third jobs.

From what I’m observing, third level institutions are also still teaching Human Resources the way it was taught ten, twenty plus years ago and are not equipping students with the new skill sets needed for where the HR profession is moving. I would also argue the likes of CIPD and SHRM are behind the curve here and are not doing enough for their members given the pace of change coming and how fast it’s approaching. (Again – walk around Unleash / HR Tech etc and you’ll see what I mean.)

What we saw in Paris.

There were hundreds of companies exhibiting, from established vendors to early stage start-ups. Around 2 out of every 3 companies is leading with an AI solution. About one in four are AI native, where AI is the product itself.

Where the AI clusters are happening says a lot. Hiring, with AI interviewers screening every candidate. (Hello Chick Bongo!)

Also – sidenote. Certain US speakers were very bullish on interviewing / screening candidates with AI robots. I noticed a distinct lack of enthusiasm for this amongst the predominantly European attendees. People still want to speak with people and I don’t see that changing for a while yet.

Back to Paris. None of the AI native companies were in core HCM or payroll. Workday, SAP SuccessFactors, Oracle, HiBob and Personio are all adding AI agents on top of their system of record. A few vendors we spoke to were at pains to tell me “this is an absolute gamechanger Cian”, but when we probed, dug a little further, we could see the dashboard, home page looks sharper but the underlying application still looks the same.

So the core system of record is not going anywhere. What we observed last week is that everyone wants to plug into it, and that’s where HR should start getting concerned.

The data question. The data in question.

Riina Hellström wrote a brilliant piece here on LinkedIn last week called “HR AI Warning – The biggest AI Flag I brought home from Unleash” – if you haven’t read it, please do so. Vendor MCP connections (Model Context Protocol) is how Anthropic, OpenAI or Microsoft Copilot pull data out of a HRIS while respecting your access rights. A manager will only see what they would see if logged in.

Fair enough. But once a manager pulls performance, absence and learning data into their AI tool, adds their own notes or observations (i.e. Cian is abrasive and hard to work with and doesn’t get on with his fellow team members), then saves this output into a new file, SharePoint or a download folder, this is a new employee record that now sits outside of the HRIS, and nobody else outside the manager knows about it or has access to it. Under GDPR, the employee is entitled to ask to see this.

Another post I spotted was from the brilliant Sharon O’Dea, who shared a story from her first day at Unleash last week. She asked an AI platform how many people work for the company and was then told (2 mins later) she wasn’t authorised to know something that’s actually in the publicly available annual report.

Almost everything she observed last week is built to make individuals more productive, but very little is being built for the organisation as a whole. Does it make sense if 2K employees have their own AI agents, with 2K personalised versions of working life, but nobody is looking after what’s being shared?

So HR now has two design problems landing at once. Agents are too loose with personal data, and they can know everything about each employee but nothing about the company. Shared knowledge, culture and how the organisation fits together have always been HR’s job. Agentic AI is making that job much harder, but equally more important at the same time!

I didn’t see / find any vendors last week that leads with governing employee data once an agent has access to it.

HR has an image problem.

Last week Josh Bersin made a prudent observation at a roundtable workshop about HR. He said “HR is great at optimising for processes that are sometimes not needed!” Think about that for a minute.

There are lots of articles out there writing about HR not getting their seat at the top table, they’re only there to serve the interest of the business and not employees, they don’t understand the P&L, only focussed on compliance, they live in an ivory tower, the fun police, yada, yada, yada! As a recovering HR practitioner myself, I do agree with some of that sentiment, but nobody seems to be looking at the bigger problem coming down the tracks for HR.

I spent nearly 20 years as a HR practitioner before moving into product and then advisory. All of the Nova team come from HR and recruitment backgrounds, and most of the people in my HR network would say they got into HR in the first place because they like people, or working with people, or helping people. Very few people get into HR because they love designing data permissions or building workflows.

You like people? Or you like People, Process and Data? Because that’s what working in HR is going to be! That might have worked when HR technology is a system you buy every 5 years or so and someone on the team who is ‘technical’ is stuck minding it. It’s clear from last week that model will soon end.

Brett Ungashick came back from HR Tech in Las Vegas last week with a story about a HRIS leader with a 6K person business. His team of 8 has a very clear mandate. Everyone has to be able to build agents and automate their tasks, or their roles will be fractionalised. This is what is coming for HR professionals. And I don’t think they’ve woken up to this yet.

This is the job HR is already being asked to do in some companies, and the job HR will be expected to do in all companies in the next 5 years. Yes. I’m predicting this sea change will happen for most companies within the next 5 years! (Except public sector of course – they’ll keep going as is for another decade at least regardless of what change happens.)

So when a talented data analyst or product manager is thinking about their next career move, HR is not making the list. Most people still see HR as the policy and process corner, or the place you go (or are sent to) when things go south. This is a big problem for HR.

The time challenge.

The thing we hear from the brilliant HRDs we work with, again and again, is around time constraints. “I just don’t have the time or bandwidth for a digital transformation, but I know I have to do it.”

And they’re right. In any given quarter the HR Team is running a calibration review, a talent review and a compliance audit. They’re ramping up in one part of the business, in the middle of pay talks or are running a redundancy programme in another (Hello Microsoft HR Team – you’ve a busy few months ahead). Sometimes a merger will land unannounced. And because HR Teams are not traditionally focussed on the HR tech stack and can’t see the wood for the trees, the HRIS will get managed in whatever time is left, which is usually none.

Unleash showcased many HR vendors who purport to have the answer, but those vendors are being run by very technically astute people whose salespeople (who have never worked in HR) are primed to sell the best version of that technical product (in most cases) to a non-technical HR buyer who is a cost centre, risk averse and scared of buying the wrong HR application. And soon that same HR Team will be asked (tasked?) to redesign how HR itself is structured, decide which % of their work is going to agents and will then be asked to then govern the data those agents touch. Nobody (from what I can see) is making any time to prepare for this.

So businesses are announcing an AI strategy, vendors are rocking up with agents priced on the company’s headcount, and HR is expected to sign off on a redesign of its own function between two review cycles? This to me is bonkers. If HR doesn’t have the capacity to lead that work, then inevitably IT and/or Finance will be asked to step in and lead it for them. And that also impacts HR’s credibility, ability to maintain its destiny etc etc.

Are companies ready for this bold new agent-led HR future?

The evidence would say no. Plus I would also say no based on what we’re seeing in the market.

McKinsey & Company, in their excellent State of AI survey in August, found 39% of respondents expect AI-related cuts to total employment in the next year. Only 14% of AI-using organisations said AI actually reduced their workforce in the last one. Tom Davenport and Laks Srinivasan made the same point in Harvard Business Review in January. Companies are laying people off on the strength of AI’s potential before it has proven its potential.

McKinsey also estimate that 90% of organisations still run an industrial age operating model. In the same research, they found a team of 2 to 5 people can already supervise 50 to 100 special agents. Those few people need to understand the work, the data and the risk. In most HR / People functions that skillset does not yet exist.

Meanwhile research from BetterUp and Stanford University, published in HBR, found 41% of workers have received AI “workslop”: polished output with zero substance behind it. I’d wager that number is higher.

Oliver Kasper summed this up in his post after the excellent Stacia Garr session in Unleash last week: “AI usage alone tells us very little about success!” Another observation from last week came from Oli Meager, who said work design was the conversation in Paris, and that Europe now feels 3 months behind the US instead of 12 to 18. I agree with him on the conversation, but I’m less convinced HR Teams are being built / prepared to do the redesigning.

My three predictions for the next five years.

1. The person who governs people data across AI becomes the most important hire in AI. Every MCP connection and every agent pulling from your HRIS creates records HR doesn’t control. The EU AI Act already treats AI used in hiring, promotion and performance decisions as high risk. I expect a major GDPR case over a manager’s AI-generated notes on an employee before 2030 (take note, Adrian Twomey), and this will motivate boards and SLTs (Sr leadership teams) faster than any policy document. The companies that get ahead of this will have a named owner for people data across every agent, sitting in HR with a direct line to IT and Legal (where appropriate).
2. The traditional route into HR closes, and the next generation of HR leaders will emerge from outside HR. The entry level work that trained most of us hard working HR folk (onboarding admin, contracts, employee queries, data entry) is exactly what the AI HR point solutions we saw in Unleash are designed to do. When that work inevitably goes away, so do the apprenticeships and grounding work for early stage HR careers. McKinsey found demand for AI fluency in US job postings has grown sevenfold in two years. HR is going to have to recruit from data, product, operations and finance, and it will only win those people if it changes how the function is seen. This won’t be an easy task. That is ultimately the image problem HR needs to tackle, and should be top of every HR Leader’s agenda now.
3. Budget moves from HR headcount to platform and services, and the big platforms are going to get bigger. Brett Ungashick is correct. Vendors are going after labour budgets, not just software budgets. We are already hearing this from the sales teams in the vendors we speak with. The platforms that own the record, Workday, HiBob, Dayforce etc, will sell agents and managed services priced against the headcount they (say) they save you.

A lot of the AI native start-ups on the Paris floor will be absorbed (acqui-hired) by those same vendors over the coming years. Or just squeezed out entirely. For the HR leaders reading this, the biggest commercial decision of the next five years is how much of your operating model you hand to one vendor, and on whose terms? HR Teams with no time to own that decision will find it will be made for them by IT or in some cases Finance, and that is not an ideal outcome either.

At Nova, our north star has always been to boost the credibility of HR in the eyes of the business. Agentic AI is the biggest chance HR has ever had to do just that. It is also the fastest way to lose credibility if the function isn’t set up to own the data, the agents and the outcomes. Most HR teams we meet don’t necessarily need another HR application. They just need the time and support to redesign the function while they’re running it.

Share on LinkedIn →First published on LinkedIn →