Companies aren’t debating whether to invest in AI anymore – look at the headlines and you’ll see that widespread adoption is well and truly here. What they ARE debating now is the only question that now matters: is any of it paying off?
That’s the paradox at the heart of modern business – companies are pouring billions into AI, yet Roland Berger research shows nearly 90% are still seeing no meaningful financial return. The problem is that too many leaders are treating AI as something you buy, not something you build and run as a capability – so spend keeps rising, expectations keep rising, and value lags. That is the AI value gap. Before organisations can close it, they need to admit why it exists. And 99% of the time, it’s down to patience, governance, closely tracking returns, and maintaining our human expertise.
Speed is not the answer
Our research shows that when companies rush AI into production, 68% of projects stall.
Top performers turn deployment into financial return quickly, but importantly, effectively – for this group, 62% hit financial targets on time. The weakest performers are those that just absorb cost without deploying properly – only 9% of this group hit their financial targets on time.
For UK businesses, it is worse. Britain shows the highest share of stalled AI programmes, higher than the US and France, and well behind Japan. The gap is not talent or tooling. It is how AI is funded, governed, measured, and embedded in a business’ operating model.
For businesses to successfully launch at pace, it is crucial that they have carefully considered and designed what happens after their AI tools go live – and even more crucial that they make it mean something for their business and their employees.
Most companies are flying blind
Most companies still haven’t got to grips with how to best leverage AI. The pattern is clear: 63% of businesses rely on one-off metrics or gut instinct rather than continuous, data-led monitoring, and only one in four tracks AI returns automatically in real time. That means many cannot tell what is creating value and what is burning cash, so both continue to be funded.
The answer is not prettier dashboards, but treating AI as a long-term asset: continuously measuring ROI, aligning leadership around a small number of enterprise priorities, and embedding governance into the platform so teams can focus on outcomes rather than paperwork. AI control towers such as central dashboards with automated KPI tracking can give leaders a real-time view of revenue, margin, retention and innovation speed.
The real opportunity is to stop obsessing over short-term savings and start building a business that compounds value.
Stuck in pilot mode
Many organisations have bought impressive AI and bolted it onto yesterday’s systems. It may demo well in isolation, but it cannot support the demands of everyday operations.
Too often, AI is still treated as a collection of discrete pilots – experimental, tightly bounded and safely contained. Escaping pilot mode requires more than new tools; it means rewiring funding models and governance, investing in shared platforms, adopting common data standards and putting controls in place that enable scale rather than trap AI in silos. When AI becomes part of the operating infrastructure, it starts delivering real operating results.
Governance isn’t the enemy
When clear ownership, sensible guardrails and automated checks are built in from the start, teams can move quickly and scale with confidence. Conversely, when governance is layered on later through manual approvals and endless sign-offs, it creates delays and encourages workarounds.
Getting this right requires genuine leadership alignment: the CEO should be focused on enterprise value, the CTO or CIO creating the data foundations and visibility tools, and the CFO backing the shared platforms that make progress possible.
The role of human expertise
Two thirds of companies struggling to make progress with AI cite a lack of technical know-how amongst their employees and talent as a major hurdle. Yet real competitive advantage still lies in human judgment, relationships and institutional memory – much of which sits in people’s heads rather than in usable data. AI cannot learn from knowledge that has never been captured.
Before scaling AI, organisations need to be honest about what only their people know, then invest in preserving, codifying and protecting that expertise.
Running a Ferrari with a go-kart engine
In Japan, companies are seeing concrete returns from AI at more than four times the rate of many European businesses, showing that the leaders are already pulling away.
The lesson is straightforward: most organisations have learned how to deploy AI, but far fewer have learned how to make it pay. As the technology becomes cheaper and more widely available, the real competitive edge will come not from owning the tools, but from extracting value from them.
You cannot buy Ferrari technology and run it on a go-kart engine.
About the Study
The study is based on a survey of 203 executives with technology-related mandates and direct oversight of or involvement in AI initiatives. The sample covers Europe, the USA, and Japan with the following geographic distribution: DACH region (Germany, Austria, Switzerland) 35%, France 15%, United Kingdom 15%, USA 20%, and Japan 15%. Respondents represent various industries, including: consumer goods, retail and logistics; healthcare and energy; financial services; industry and automotive; as well as technology, media, and telecommunications.
About Roland Berger
Roland Berger is one of the world’s leading strategy consultancies with a wide-ranging service portfolio for all relevant industries and business functions. Founded in 1967, Roland Berger is headquartered in Munich. Renowned for its expertise in transformation, innovation across all industries and performance improvement, the consultancy has set itself the goal of embedding sustainability in all its projects. Roland Berger generated revenues of around 1 billion euros in 2024.
