Depending on what study you believe, according to recent research anywhere between 70% and 90% of UK businesses are now using artificial intelligence (AI). On the face of it, that sounds like good progress: AI has moved from hype to becoming a habitual part of the way most organisations work.
But scratch the surface and there is a fundamental question we need to answer: are we using AI for anything genuinely productive, or just ticking a box that says ‘we’re doing AI’? The reality is, a lot of organisations are using the technology in some way, or incorporating it into their products and workflows, but in too many cases they are doing it for its own sake rather than with a measurable outcome in mind.
When ‘use more AI’ doesn’t work
Amazon’s now-scrapped AI leaderboard is a case in point. The idea was simple: encourage employees to embrace the technology by ranking them on how often they used it. But more prompts does not automatically mean better work or actual impact – in fact, it can just drive people to use AI regardless of whether it actually helps.
That’s the AI trap many organisations are falling into. They jump to adoption without first asking what problems are slowing them down, and where AI can genuinely help address that challenge.
Beginning from the position of ‘we need to use AI’ and you always risk falling into unproductive usage and even novelty. But, focusing on the need to improve productivity, quality, or customer experience puts you in a much better place that may lead to actual impact.
Start with bottlenecks, not tools
From our experience, the most effective AI programmes begin by looking at where bottlenecks lie – repetitive, time‑consuming tasks that take time and resources without delivering much value. In the case of software development, that might be: writing test cases, first‑pass code reviews on pull requests, or documenting meetings and actions.
When AI takes on those tasks, not only do engineers get time back to focus on the work that actually needs human judgement – whether it is designing systems, solving complex problems, or thinking about architecture – you should also see faster pull‑request turnaround, fewer bugs, and shorter test cycles.
The same can be applied in customer support. Many organisations rushed to add AI chatbots because they would ease the pressure on overstretched teams. But, in many cases what has actually happened is customers being forced through extra steps just to reach a human, with drop‑off rates spiking and the underlying issues remaining.
Define success before you deploy
Rather than jumping straight to the solution, identifying the problem is the better place to start – and more often than not, that means speaking to people at the front of delivering these services. They will be able to say where they are most over-stretched, which tasks may be repetitive but are required, and where time can be freed up to focus on where they can make the biggest impact.
From that, the answer is rarely another chatbot or automating what seems like the most obvious function. Instead it tends to be answers such as AI‑generated call notes, summarised tickets, or suggested responses – tools that quietly remove friction, but are not necessarily big, exciting projects. And, over time, they should translate into measurable success in the form of reduced waiting times, higher customer retention, shorter development cycles, or fewer security vulnerabilities.
The measure of success should be clear and agreed upfront. If you introduce an AI chatbot and your customer services drop‑off rates skyrocket, then that is completely counterproductive, even if the team has ‘adopted AI’. On the other hand, if AI transcription cuts customer call time by an hour per day, that’s a measurable win.
Why culture is key
There’s another layer that matters just as much as AI tools and the metrics used to measure their use, and that is culture. Many people are understandably anxious about AI. They worry it will replace their job, or expose their mistakes, or be used to quietly judge their performance. If you want AI to be used well, you have to address those concerns head‑on.
In practical terms, that means making it clear, from the top down, that AI is there to support productivity, not to quietly automate people out of a role. It also means creating space for experimentation, where individuals can try AI, learn from the experience, and share it without fear.
In that respect, embracing AI has to be treated as a collective endeavour. Managers can bring ideas to the group, but experimentation also has to be on an individual basis. On top of that, frameworks need to be set out to ensure use is consistent and outcomes are measurable. The sweet spot is combining empowering individuals with strategic direction from above.
Is it the right tool – and is it worth it?
Another common misstep is treating all AI tools as interchangeable. Different platforms are good at different tasks – some are great for code review and pull requests; others are better at documentation, test cases, or workflow design; while some are better suited to creative writing or building presentations. Don’t put the cart before the horse. Ask what problem you are trying to solve, and identify which tool is best suited to that.
There is also a harder question that responsible organisations are beginning to ask:
Is this use case worth it? If an AI chatbot consumes huge compute resources, drives up costs, uses water‑intensive data centres, and doesn’t actually improve customer experience, then it may not be worth the environmental impact or the financial risk.
AI is not a sticking plaster for every problem. Sometimes, the right answer is to not use it.
Using AI well
For many organisations, adopting AI the right way is about a mindset shift rather than finding the right tool or using the technology at all. Instead of asking how you can use more AI, it is better to ask where it can genuinely help you achieve impact.
From that point, you can more clearly define success in measurable terms, build a culture where experimentation is encouraged and outcomes matter more than usage, and stay honest about costs, risks, and impact.
AI can absolutely improve productivity, quality and customer experience. But only if we treat it as a tool in the service of delivering an outcome that actually matters – not as an outcome in itself.
