The AI Challenge Facing Today's Leaders

Organisations face difficult choices as they seek to create value from AI. Headcount reduction has dominated. Yet one in three AI-driven layoffs is already being reversed. The failure is in leadership judgement, not technology deployment.

Following our recent webinar, Deliberately Human Leadership in the Age of AI, one theme emerged consistently: while organisations are investing heavily in AI capabilities, many are underestimating the leadership challenges that determine whether those investments succeed or fail. In recent conversations with HR teams in large organisations, I’ve found that discussions about AI adoption create considerable tension. On one side, they are being asked to lead the workforce through the biggest shift in how work gets done in a generation. On the other, they are watching AI reshape their own function in real time. It’s a complex scenario. And yet the numbers are making it harder to ignore.

In the first half of 2026, AI became the leading stated cause of layoffs across US tech, driving well over 100,000 announced job cuts across major employers. Over the same period, staffing firm Robert Half found that nearly one in three US hiring managers who had cut a role primarily because of AI has already rehired for the same job or something close to it. Separate research from workforce planning firm Orgvue put the share of business leaders who now regret AI-driven layoffs at 55%.

This is a Management Failure

That is a striking set of numbers. The obvious reading is that AI is not as capable as its more enthusiastic advocates claimed. That may well be true. But it is also the least interesting interpretation. A more useful reading is that these were management failures dressed up as technology decisions. Nvidia’s Jensen Huang, no critic of AI, put it more bluntly earlier this year when he called executives who blame layoffs on AI “lazy”. His point was operational rather than philosophical: if you have already deployed AI to the point where you can eliminate roles, but the roles keep coming back, the failure sits with the people who defined the job in the first place.

Klarna is the case most people know. In February 2024, the fintech announced with OpenAI that its AI assistant was doing the work of around 700 customer service agents, handling two-thirds of chats and cutting resolution times from about 11 minutes to under 2. Fifteen months later, in May 2025, CEO Sebastian Siemiatkowski told Bloomberg the company was hiring humans back. In his own words, cost had become too dominant a factor, and “what you end up having is lower quality”.

The AI had not failed. What had failed was the definition of the job it was asked to do. The dashboard measured throughput; the business needed customer satisfaction. Nobody noticed the gap until the customers did.

Why HR Sits at the Sharp End

This is where the story stops being about customer service and starts being about HR. Because right now, HR and L&D functions across organisations are being asked to lead a transition that also lands directly on them. Much of the day-to-day work of these functions, including routine employee queries, onboarding logistics, standard learning content, and first-line policy questions, is exactly what current AI does well. This places HR in the eye of the AI storm.

We can already see some of the implications of this. At IBM, its AskHR system reports handling around 94% of routine HR queries. IBM’s CHRO Nickle LaMoreaux describes the effect as freeing HR staff for “higher-value work that people are more engaged in”. But the deeper reading, as Inc.’s Heather Wilde put it recently, is more pointed: “the routine work was never where your people’s value lived. If you cut heads based on what AI can do on an average day, you’ll pay for it on the hard days.” Her line at a Charter AI Summit in New York is worth considering: “If we don’t continue to invest in entry-level hires, what happens in three to five years? There’s no pipeline; the well simply dries up”. That is a leadership judgement failure, not a technology decision.

HR is the leading indicator here, but the pattern will hit every function. What starts as a productivity opportunity becomes a capability gap: the sort of gap that Ford’s decision to rehire more than 350 veteran engineers, to catch quality issues its automated systems had missed, makes concrete.

In my recent book, “Making AI Work for Britain”, I review such AI adoption issues and their consequences in the context of the UK’s AI strategy directions. I argue that most organisations are stuck in what I call pilot purgatory. A great deal of activity, very little value, because leaders are buying AI capability faster than they are building the organisational capability to use it well. The rehiring wave is what pilot purgatory looks like once the accounting catches up with reality.

HR and L&D leaders occupy a particular version of this problem that no other function faces in quite the same way. They are being asked to lead the workforce through an AI transition that is simultaneously reshaping their own roles. Every conversation about workforce readiness is also, in addition, a conversation about their own function. It is one thing to help the finance or operations group rethink their work. It is another to do that while your own job description is being rewritten in front of you.

Judgement Not Tools

There is a way through, and it is not to pretend the tension isn’t there. The functions that will emerge stronger from this cycle are the ones that redesign themselves first, visibly, before offering to redesign anyone else. That means building an effective smart-buyer capability to strengthen the organisation’s ability to specify, procure and manage AI systems well, not simply use them. It means shifting learning investment away from tool training toward judgement training, because prompt engineering is a two-week skill and knowing when to override a model is a long-term behaviour change. And it means being willing to kill pilots that will never scale, to free the capacity, and focus the courage, for the ones that will.

The organisations pulling ahead in AI are not the ones with the most tools. They are the ones with the leadership judgement to know which tools to buy, which to kill, and when to walk away. The rehiring numbers are telling us something important, but not what the headlines are saying. They are telling us that the constraint on AI value is not the technology. It is the capability of the people making the technology decisions. A challenge that every leadership team must urgently address to move forward with AI.

None of this is a call to slow AI adoption – it’s a call to lead it more deliberately. The organisations getting this right are treating judgement as a capability to be built, not a trait to be hoped for. For a deeper look at what that capability requires in practice, read The Oxford Group’s whitepaper, Leading in an AI World, which explores how leaders can develop the judgement AI adoption demands.