KPMG: The real AI bottleneck isn’t tech, it’s humans
The challenge with AI is no longer about whether the tech works, but whether the people and processes around it are ready, claimed business and tech leaders at KPMG’s panel at Viva Tech Paris.
“We see the potential”, said Nicole Buttner, chief executive of Melantec and general secretary of Germany’s MPP party. “But it’s like this uncomfortable transitional period where the promise of generative and conversational AI has not fully hit organisations”.
Europe, once seen as a cautious player in the AI race, is now grappling with hard operational realities of siloed systems, fragmented data strategies, and clear change-resistant work cultures.
And as the UK attempts to lead the way, the warning signs persist.
Tech isn’t the problem, trust is
Despite significant investment in AI innovation, returns remain elusive outside of narrow machine learning applications.
“The real, scaled ROI today is in traditional areas, computer vision, classic NLP”, Buttner said. “Generative AI? We are simply not there yet in terms of enterprise impact”.
Ashish Badan, one of KPMG’s chief tech officers, likened the situation to “owning a sleek new Porsche, fully equipped…but not having a licence.”
The issue, he told Viva Tech attendees, lies in organisational infrastructure, not GPUs or models. “You need governance, systems, and people ready to make use of the tech”.
This mirrors a broader challenge facing UK businesses, many of which are running enthusiastic and ambitious pilot schemes but struggling to move beyond experimentation.
For robotics pioneer Dr Amit Andre, the convergence of AI and robotics – what he dubbed “physical AI” – marks a new phase of capability. “But with that comes a deeper question: are we applying it where we need it?”
He challenges the rush to deploy AI for the sake of it, underscoring the risk of performative innovation – particularly acute in publicly funded or high-stakes sectors like UK healthcare and government.
People rather than processors
The panellists repeatedly returned to the argument that it is not the chips or algorithms holding back AI, but human application.
Governance, incentives and cultural buy-in are the real bottlenecks.
Buttner noted: “Many businesses and governments have the same problem of implementing different, incompatible stacks”.
The UK faces a similar decentralisation, with numerous NHS AI pilots underway, yet few have been integrated at scale.
Without coordinated digital infrastructure designed to enforceable standards, even effective solutions may risk being isolated.
“Right now, organisations are feeling the coordination problem, especially as they deploy AI agents”, Buttner described.