Businesses can now buy AI through productivity tools, language models, analytics platforms, automation software and specialised applications. Buying access is straightforward. Establishing where it creates measurable value takes more work.
The commercial return on that investment is, by most honest assessments, deeply uneven.
Some businesses have transformed specific operational capabilities using AI. They have reduced costs, improved quality, created competitive advantages, and scaled value in ways that would have been impossible without the technology.
A much larger number of businesses have spent clear money on tools that are used by some people sometimes, have not changed how the organisation fundamentally operates, and are quietly becoming another line item in the IT budget that gets renewed because cancellation is harder than continuation.
The tool is not the differentiating factor. The strategy — or its absence — is.
What a strategy actually requires
An AI strategy is not a statement of intention or a list of tools. It is a set of explicit choices about where AI will change specific things the business does, by how much, with what investment, owned by whom.
Those choices require a level of specificity that most businesses have not committed to. “We are implementing AI to improve productivity” is not a strategy. It is a mission statement for an initiative that has not yet been designed. A strategy says: we are using AI to reduce the time our proposal team spends on first-draft commercial documents by 60%, with named ownership, defined success criteria, a six-month timeline, and a workflow redesign that makes the tool a standard part of how proposals are produced — not an optional shortcut.
The specificity is not pedantic. It is what separates investment that produces returns from investment that produces activity.
McKinsey’s 2025 global survey of AI adoption found that organisations generating the strongest value from AI were distinguished not primarily by technical sophistication but by the clarity with which they defined use cases, the discipline with which they redesigned workflows, and the seniority of the leadership ownership behind their AI initiatives. These are management and governance characteristics, not technology characteristics.
Why tools without integration plateau quickly
The pattern of AI investment that stalls is remarkably consistent. An organisation invests in a capable tool. Early adopters — usually the most technically confident and curious members of the relevant team — start using it. They find it useful. They report time savings. Leadership is encouraged.
Then adoption flattens. The early adopters continue using the tool. The broader team uses it occasionally, when they remember, or when a specific need makes the effort worthwhile. But it is not part of how the team routinely operates. It sits alongside workflows rather than inside them.
At this point, the investment is producing some value — but less than it could. And the business often responds to this by looking for a better tool, a more user-friendly interface, or a more aggressive adoption campaign.
The diagnosis is usually wrong. The problem is not the tool. The problem is that the workflow was not redesigned to make the tool a necessary component of how work gets done. Optional adoption follows naturally from optional integration. Consistent value requires consistent use, and consistent use requires the tool to be embedded in the process rather than available alongside it.
DSIT’s AI Adoption Research found that 80% of AI-using UK businesses were using their tools at least weekly. Frequency alone does not establish how much value a tool creates. That depends on the task, the time saved, the quality of the result and the cost of checking it. The question is not whether the tool is being used. It is whether it has meaningfully changed how work is done.
The use-case problem
ONS data identifies use-case identification as the single most common barrier to AI adoption — cited by 39% of UK firms. The significance of this finding is not that businesses are bad at identifying use cases. It is that it reveals where the actual problem sits: not downstream in technology or cost, but upstream in strategic decision-making.
A business that does not have clear answers to the question of which specific operational problems AI should solve has not yet made the choices that allow an AI investment to be deployed effectively. Everything that follows — tool selection, vendor engagement, implementation, adoption — is building on a foundation that is not yet established.
The businesses that invest in use-case clarity first — before evaluating tools, before engaging vendors, before building business cases for specific platforms — consistently produce better outcomes from AI investment. This is not a slow path. It is a faster one, because it avoids the cycle of tool acquisition, shallow adoption, disappointment, and reinvestment that characterises the more common approach.
What strategy looks like in practice
An AI strategy does not need to be comprehensive to be useful. It needs to answer a small number of questions with enough specificity to guide investment and accountability.
Where in the business do we have operational problems that AI could solve — not in theory, but specifically, with identifiable costs and measurable outcomes? Of those problems, which represent the highest-value opportunities relative to the effort required to realise them? What changes to workflows and operating processes would be required for AI to provide that value — and are those changes achievable? Who is accountable for the commercial outcome, and what does success look like in 90 days, six months, and a year?
Answering those questions before investing in tools produces a very different investment decision from the one that most businesses make. It also produces a very different implementation plan — one focused on operational change rather than technology deployment.