If you ask HR professionals today what they are doing with AI, you will probably get a wide variety of answers. Some are using ChatGPT to write content, others are experimenting with Copilot, recruitment teams are using AI for matching or screening, and somewhere there may already be a first project involving a chatbot or an AI agent.

There is nothing wrong with that. Quite the opposite. Experimenting is probably one of the best ways to discover what AI can actually mean for HR.

But I increasingly wonder what happens when we ask a second question: what is the strategy behind all these AI initiatives?

I suspect the answer becomes much less clear.

We have been very quick to apply AI to individual HR problems. Writing job descriptions takes too much time, so we use generative AI. Recruiters spend too much time searching, so we introduce AI matching. Candidates need faster responses, so we look at chatbots or agents. There is too much administration in the process, so we try to automate it.

All useful. But together, they still don’t necessarily form a strategy.

An AI strategy for HR should, in my view, not even start with AI. It should start with HR itself. What do we want to do better as an organisation over the next few years? Do we want to find people faster? Do we want to reach better candidates? Do we want to give recruiters more time for the genuinely human part of their job? Do we want to better understand the skills we already have within the organisation? Do we want to encourage internal mobility? Or do we finally want to get more value from the enormous amount of candidate and employee data we have collected over the years?

Only when those questions are clear does it become interesting to look at where AI can play a role.

And I think this is an important distinction from the way we often look at AI today. AI is not simply another tool to add to the HR technology landscape. It can actually connect different parts of that landscape.

Take recruitment. A candidate applies and AI can interpret and structure the CV. The profile can then automatically be matched against the vacancy. Perhaps some of the information is missing or outdated and can first be checked or enriched. An AI agent can then ask additional questions or find out whether the candidate is still interested. Eventually, the recruiter doesn’t simply receive a pile of applications, but a much better prepared overview of relevant candidates.

That starts to look much more like a strategy than a collection of separate AI tools.

The same applies to the existing candidate database. Many organisations have spent years building huge databases of profiles in their ATS or CRM. A lot of money and time has gone into creating those databases. Yet a significant part of that information is hardly used anymore because recruiters cannot find the right candidates, profiles are incomplete, or the information has simply become outdated.

The strategic question then isn’t: which new AI tool should we buy? The better question is: how can we use AI to unlock the value of what we already have?

This also makes it clear that AI isn’t only about automation. The discussion today is moving very quickly towards AI agents: which tasks can we hand over to AI? But at least as important is how AI can help us make better decisions, create better matches and work with better data.

Because ultimately, a simple rule still applies to AI: poor or outdated data rarely produces good results. An intelligent agent working with incorrect information will continue to use incorrect information. A good matching algorithm can only perform optimally when the information it uses is sufficiently rich, relevant and up to date.

That is why I think HR should spend a little less time thinking about individual AI features and a little more time thinking about the bigger picture.

What do we want to do better? Which processes could fundamentally change? What data do we need to make that happen? Which tasks can AI take over? Which decisions can AI support? And, equally important, where do we explicitly want people to remain in control?

Once you start doing that exercise, something changes. You no longer have an ATS with ChatGPT sitting next to it, plus a chatbot, a matching tool, an agent and a handful of other AI experiments.

Instead, you start building an HR environment in which data, intelligence, automation and human expertise work together.

Perhaps that is ultimately the most important question we should be asking HR today.

Not:

“What are you already doing with AI?”

But:

“If I look at everything you are doing with AI, can I see the HR strategy behind it?”

If that is difficult to explain, perhaps we don’t have an AI strategy yet.

Perhaps we simply have a lot of AI.