Updated 17 April 2026 — The 15 April Copilot Chat restriction has now been live for two days, and the initial reactions are worth pricing in. This revision also adds the DSIT January 2026 figure (just 16% of UK businesses actually use AI), the global investment-impact gap (56% of CEOs report zero financial benefit, 95% of custom pilots fail) and a note on enterprise-scale AI workforce training (Lloyds Banking Group), so the numbers SMEs hear match the numbers they should actually make decisions on.
Here's a conversation I have all the time. A business owner tells me: "We're not really an AI company. I don't think AI is relevant to us." Five minutes later, they're describing how their team spends two hours a day answering the same customer emails. Or manually typing invoice data into a spreadsheet.
They don't have an AI problem. They have a productivity problem. They just don't realise that AI is now the fastest, cheapest way to fix it.
There's a massive disconnect between what AI can actually do for a small or medium-sized business and what most business owners think it does. When people hear "AI," they picture self-driving cars and robots. That's not what I'm talking about.
I'm talking about a tool that reads your incoming emails, drafts a professional reply in your brand voice, and sends it to your team for one-click approval. Or a system that automatically reads every new contract and tags it with key dates and renewal deadlines.
These aren't futuristic. They exist right now, they run on the Microsoft infrastructure most UK businesses already have, and they pay for themselves within weeks.
The AI industry has done a terrible job of explaining what's actually useful. The conversation has been dominated by hype — generative AI, large language models, artificial general intelligence. None of that means anything to a managing director who just wants their operations team to stop drowning in admin.
The real question isn't "should we use AI?" It's "where is my team losing time on work that doesn't require human judgement?" Once you answer that, the AI conversation becomes practical instead of abstract.
Adoption has nearly doubled in two years, but only 11% of UK SMEs use technology extensively to automate operations. The headline number is moving fast; the depth of use isn't. Buying a Copilot subscription is not the same as actually changing how your team works.
It’s also worth stress-testing that 54% figure. The same month, DSIT’s January 2026 research pegged actual AI use across UK businesses at just 16%. The British Chambers of Commerce number for SMEs (35%, up from 25% in 2024) sits between the two. Three different numbers, three different methodologies — and all three broadly say the same thing: headline AI adoption is rising fast, but deep, measured, process-level use in SMEs is still the exception. That gap is the actual opportunity.
Globally, the same pattern shows up on the cost side. AI spend hit £2.52 trillion in 2026 (a 44% jump), yet 56% of CEOs report zero financial benefit and 95% of custom AI pilots fail to deliver P&L impact. Translation: a lot of businesses are buying AI. A much smaller number are actually earning back what they’re spending on it. The winners aren’t the ones with the most tools — they’re the ones with data foundations, governance and a clear "what exactly are we automating?" answer before any purchase.
Enterprise is starting to act on this. Lloyds Banking Group, for example, has committed to training every employee in AI skills by the end of 2026. That’s not a pilot — it’s a workforce bet. UK SMEs don’t need to match that scale, but the direction of travel is useful context for any owner still framing AI as optional.
On 15 April 2026 Microsoft restricted Copilot Chat inside Word, Excel, PowerPoint and OneNote for users on the free Microsoft 365 Copilot Chat tier. Full in-app Copilot now requires a paid Microsoft 365 Copilot (Premium) licence — currently around £24.70 per user per month in the UK. Many businesses who thought they "had AI" because Copilot lit up in their Office apps have discovered they didn't — and that they never built a process around it.
Two days in, we’ve seen two reactions: (1) panic purchasing of Premium licences without a clear use case, and (2) a quieter, smarter response — leaders pausing to ask "what were we actually using it for, and is a paid licence the right answer?" Sometimes it is. Sometimes the better answer is a targeted custom automation that costs less than the licence and solves one specific problem properly.
This is the clearest illustration yet of why "subscribed to AI" and "using AI" are not the same thing. You don't need AI in the abstract. You need someone to set it up against a specific problem in your business, and design the workflow around it — whether that's Copilot Premium, a custom tool, or both.
Since the Copilot Chat change, we've had more enquiries than usual from SMEs who don't want to sign up for a year of Premium licences without knowing whether the tool actually fits their work. That's the whole reason the AI Growth Lab exists.
It's a low-risk sandbox: we take one real process from your business — invoice handling, contract review, email triage, meeting follow-ups — and build a working automation in your tenant. You see it running on your actual data, with your team, for a fixed fee. No annual commitment, no platform lock-in. If it works, you scale it. If it doesn't, you've learned what doesn't fit without having bought licences for 30 people first.
Post-15 April, that matters more than it did six weeks ago — and given the 95% pilot-failure rate globally, the point of the Growth Lab is to be one of the 5%.
Start with one process. Pick the thing that costs the most time or causes the most frustration. Describe it clearly. That's your starting point.
Usually this means someone bought a tool without a clear use case. AI works when it's deployed against a specific process with clear rules and human oversight — the opposite of the 95% of pilots that fail.
A targeted AI automation typically costs less than a month of the staff time it replaces. The maths usually works out within the first month — and is usually cheaper than rolling Copilot Premium out to the whole team.
Run everything inside your own Microsoft Azure tenant. Your data doesn't leave your environment. AI API calls are processed in-memory with no training on your data. GDPR-aligned from day one.
Agentic AI (autonomous tools that execute multi-step workflows on their own) is the dominant 2026 trend, and you'll hear about it constantly this year. For most SMEs the honest answer is: not yet. Get the boring automations working first — invoices, emails, meeting actions. Once those are reliable, agentic workflows are the natural next step rather than a leap of faith.
That exercise takes ten minutes and gives you everything you need to have a productive conversation about AI — or to scope a first AI Growth Lab sprint.
Book a free 60-minute call. I'll ask about your biggest operational headaches and show you what AI can realistically do about them — no jargon, no sales pitch.
Book a Free Call