AI investment is now a practical reality for business. KPMG’s AI Pulse shows that UK organisations have committed to be investing in AI. Use cases are multiplying and agentic AI is already being introduced into operations.
That shift is happening fastest at the top end of the market, among large corporates and listed businesses. In private enterprise, adoption is progressing more slowly.
But private enterprises are not standing still.. KPMG’s Private Enterprise Barometer found that investment in technology including AI is the top investment priority amongst private enterprise leaders, cited by 77% of executives. The primary benefits of AI are seen to be efficiency and productivity gains (59%), better quality decision-making (55%) and accelerated innovation (54%). Encouragingly also, there is a clear focus on improving data management and quality – a foundational element for successful deployment of AI.
The most important question is what a sensible approach to AI looks like for a mid-market business.
How much time and money should you pour into your own bespoke implementations – and how much should you wait for increasingly implementation-ready AI solutions to become available?
The answer will depend on the business. Sectors such as technology, business services, consumer & retail and TMT may have more AI-adaptable processes and interfaces than, say, heavy manufacturing or industrial companies.
That said, most mid-market businesses should be able to generate meaningful gains from AI.
One of the chief benefits we see reported by the large clients we are working with is that AI creates capacity gains – freeing up people to spend more time on value-adding, strategic work by automating manual tasks. In smaller businesses where, for example, a Finance team may consist of just a handful of people, freeing up half a day of an individual’s time could have a proportionately much bigger impact than where there is a team of dozens.
In areas such as Finance, HR, Tax and Legal, that kind of capacity release can remove constraints and help the business pursue growth more effectively.
So why is AI implementation still moving more slowly in the mid-market? I think there are two main reasons — but both are less significant in practice than many businesses assume.
The first one is the capability gap. Private enterprises may simply not have the data scientists and architects available to them that a big bank or telco does. However, the reality is that in a smaller business, you don’t need a large (and expensive) team: probably only a couple of tech-savvy team members are required. They don’t need to be right at the cutting edge of the latest iterations either – some prior experience of working on different forms of AI like automation of processes and a good general awareness of the AI market is generally sufficient. An advantage that mid-market businesses have is that these people can be embedded straight into the business, working directly alongside the teams they’re supporting. It’s a strength of the mid-market model compared to a large corporate where teams are much bigger and geographically dispersed. In short, the capability gap is not nearly as big for private enterprises as it seems.
The second is the confidence gap. Part of that comes from capability: if a business does not feel it has the right people around it, hesitation is understandable. But confidence should also be rising for a simpler reason: the market is improving quickly
Whether it’s ChatGPT, Copilot, Gemini, Claude or another platform, it is becoming much more feasible to apply AI to common business processes and challenges without having to do large amounts of bespoke technical work.
Against that backdrop, I would encourage private enterprises to start actively exploring AI if they have not already. Even a fast-follower strategy requires groundwork.
Certain foundations will need to be in place and you’ll also need to understand the adaptations required as AI changes how things are done. This only comes by doing.
That is why early experimentation on real business problems is so important. Start small, with one or two use cases that matter. That might mean improving stock control, speeding up invoice processing, generating better customer insight or strengthening the website experience. Make it cross-functional, involve the people closest to the problem, and treat the work as a practical exercise rather than a theoretical one.
We are also seeing two practical enablers of success in client work. The first is to move beyond the familiar idea of “human in the loop” towards what we call “human in the lead.” That means placing people at the key points in a process: setting direction, making choices, resetting objectives and judging trade-offs. AI does the heavy lifting in between, but people remain in control of the workflow. That tends to work better because it reflects the role humans actually play best: bringing business context, judgement and pattern recognition. “Human in the loop” can too easily turn into people checking large volumes of AI-generated work after the fact. “Human in the lead” is more deliberate, and usually more effective.
The second is straightforward but often missed: tell the AI to check its own work. That can mean building a formal review step into the workflow, or using a second agent to test or challenge the output. It is a simple discipline, but one that can materially reduce hallucinations and improve quality.
There is a third barrier to AI deployment in the mid-market – and that is cost. It doesn’t have to be prohibitively expensive, but of course it does require some investment. However, I would argue that the cost of doing nothing is likely to be significantly higher. It is always hard to make up ground from a lagging position. Active competitors could be way ahead in just a short time. Can you afford that risk?
The capability gap is smaller than you think it is, and more and more enterprise solutions are becoming available: the time to act is now.
Discover why organisations across the UK trust KPMG to make the difference and how we can help you to do the same.
