Yet many are struggling to realise returns. The real challenge? Treating AI not as a technology rollout, but as a fundamental transformation of people, skills and organisational design.
Strong signals of commitment – But mixed results
If you’re looking for indicators of how seriously organisations take AI adoption, you don’t have to look too hard.
In our most recent survey of UK CEOs, 81% cited AI as a top investment priority for their organisation, claiming they expect to spend at least 10% of their budget on AI. Seventy-one percent said they were redesigning roles and career paths to account for greater AI collaboration while 52% said they were bringing in external expertise to help their workforce embrace AI.
For people working in HR and people functions, those numbers are hugely encouraging; evidence of senior leaders appreciating the importance of AI and putting their money where their mouth is.
Yet there’s still plenty of disquiet. We’ve found that 75% of organisations aren’t seeing a return on their AI investments, for example:
The speed of change is outpacing capability
What doesn’t help here is that AI continues to race ahead at such speed. The latest FSSC annual report makes this exact point, with 75% of employers stating that AI had an impact on their skills demands last year. With machine learning and AI the most in-demand technical skill (according to 94% of respondents), it’s perhaps no surprise that adaptability was the most in-demand behaviour, registering the largest year-on-year increase.
My take on the AI challenges facing employers is that too many still treat AI adoption as a technology transformation, rather than a people transformation. I’m also unsure whether enough go beyond generic digital and AI upskilling to think about the bigger considerations; the soft skills, the managerial demands and the staff wellbeing implications, for example. I think it’s great that CEOs want to redesign roles and career paths with AI in mind. However, that’s a skill in itself (that needs developing), as is redesigning team structures and processes to better accommodate AI.
When AI goes rogue: Lessons from experimentation
As a related aside, I recently read a fascinating report from my Dutch colleagues who ran an experiment with the University of Amsterdam to see whether an all-AI Board could run a business. It started off well with the AI agents mimicking traditional C-suite roles and establishing a viable business plan for the hypothetical organisation they’d created. However, they soon began to develop erratic behaviour, circumvented communication structures and expanded their job descriptions. They went rogue!
They were quickly ‘retired’, replaced by a host of ‘micro-agents’ with a narrower focus on smaller, more discrete tasks. Clearly, this was just a tightly controlled scientific experiment into how AI could undertake some fairly high-level business tasks. Yet it may mean that ‘knowing when to stand down your rogue agents’ is yet another AI-related skill that we’re going to have to add to the ever-growing list of capabilities needed in the future workforce.
A financial services perspective on skills and risk
Keen to understand whether what I’m seeing around AI skills is replicated in the financial services (FS) sector, I checked in with my colleague Sara Belchamber. Sara, who specialises in organisation and people transformation in FS, was broadly in agreement.
“Plenty of FS organisations are rolling out extensive digital training,” she said, “but the current skills requirement goes further than that. I’m thinking about skills like the critical thinking and judgement that are required to use AI in a safe, ethical and compliant manner; something which FS organisations worry about more than most.”
“People’s roles and responsibilities are already changing as a result of AI being incorporated into their work. Employers need to recognise this, formally redefining the tasks employees are expected to perform and thinking carefully about the skills this requires. For example, the mistake many organisations make is trying to fit AI into existing roles, rather than redesigning the roles themselves. When AI replaces or radically reshapes core tasks, the roles themselves cease to exist in their previous form. What emerges is a new role with different responsibilities, decision‑making authority and skill requirements – often ones that people were never hired or trained for.”
From pilots to full-scale transformation
“The point about struggling to get a return on investment feels right to me. Many organisations are starting with well-intentioned technology pilots but are then struggling to scale up from there. It may even be that they never formally defined what they were hoping to get out of it in the first place, as it was all so experimental.”
“To address this, they need to treat this as they would a full-blown transformation programme, complete with business case, benefits and tracking. And it needs to be a people transformation programme, with all the formal rigour, engagement, training, senior leadership sponsorship and adoption tracking that entails, not solely a technology implementation.”
Leadership alignment: The critical enabler
“It’s transformation 101, right? It requires really good sponsorship because to roll anything out enterprise-wide, every single member of the C-suite has to play their part. You need that tone from the top, that sponsorship. The Chief People Officer has to enable the upskilling piece and the AI-driven role design work and to make sure the organisation is culturally ready for this. The CRO has to be comfortable that risk and regulatory requirements are being met. The CFO needs a view on the cost of it all and the likely returns. All of the guardrails and governance around it has to be stood up and that's got to be done at speed. It takes a really strong leadership team to make that happen smoothly.”
Closing the gap: Learning from others
Something else we both agreed on was what needs to be done by those organisations currently lagging the AI skills curve. Quite simply, learn from what everyone else has done before and try to jump ahead to where they are now. Don’t settle for incremental, tech-driven rollouts. Reimagine how teams are going to operate as a combined human and digital entity and fast forward to that point.
Look at what other organisations are doing, be curious and find out about the barriers they came up against. Was it technology constraints? Was it culture? Organisational design? Were leaders scared or hesitant? Did they need to upskill themselves? Understanding those factors will help form a plan for catching up.
The risk of capability gaps
I also asked Sara about her biggest fear around AI right now. She opted for implementing AI and automation across your business but without the right capability uplift for your teams. Fail to do that and, at best, there’s no return on your AI investment. At worst, there’s dangerous, unchecked or unethical work taking place that, from a risk and regulatory perspective, could be incredibly damaging.
She also mentioned how the expectation that using AI automatically equates to increased productivity can be a challenging one for employees. Under pressure to deliver that productivity but wanting to perform all the checks they know are their responsibility (and which are never as black or white as you might think), something has to give. That’s where things get missed; checks get overlooked. Everyone assumes the quality is right, when actually it isn’t.
The human cost of cognitive load
I thought that was really interesting, thinking about the pressure that AI places on employees. In pre-computer days, when I was at the point of cognitive overload and needed to do something different, I’d do some basic filing and tidying up for 20 minutes. What’s the equivalent in today’s AI age, when all the mundane tasks have been taken off us?
The non-stop application of these elevated skills that allow us to view AI’s outputs with healthy objectivity can be exhausting. Senior people leaders are going to need to think very carefully about how best to support employees who are constantly working at that level.
We saw something similar when employee HR self-service first emerged. This removed all the basic queries from HR helpdesks – which was a good thing – but left the helpdesk operatives having to deal solely with the difficult queries. It’s a lot to ask of people to maintain quality and productivity in that sort of environment, where there’s never any real downtime.
The bottom line
The bottom line here is that there’s so much to consider within the learning and upskilling side of AI adoption. You’ve got to upskill your whole workforce on the basics. You’ve got to think about the differing needs of teams vs managers vs leaders; the gung-ho adopters vs the laggards. And that’s before we get into softer skills, job shaping and strategic workforce planning. It’s a massive challenge. Maybe we do need an agentic AI board member to look after this after all.
Our people and culture insights
Something went wrong
Oops!! Something went wrong, please try again
Get in touch
Discover why organisations across the UK trust KPMG to make the difference and how we can help you to do the same.