Here's why.

Picture an ops manager on a Monday morning in April. Her business has had Microsoft Copilot for six months. The licence was approved, IT got it set up, and she sat through a 45-minute introductory webinar in January. And yet she's still manually copying data between spreadsheets at 8:30am, still reading through a 40-email chain to find a decision that was made three weeks ago, still doing the same things she was doing before the subscription started.

She's not alone. According to a 2026 benchmark report from Spicy Advisory, only 31% of UK organisations report a positive return on their AI investments, despite 85% to 91% of businesses increasing their AI budgets this year. Meanwhile, the UK government's own research from DSIT (the Department for Science, Innovation and Technology) found that just 16% of UK businesses are actively using AI in any meaningful way.

Organisations are spending more on AI tools than ever. Most of them aren't getting the results they expected. And the reason almost always comes down to the same thing.

The tool isn't the problem

There's a common assumption in how businesses approach AI: get the right software, and the efficiency gains follow. Buy Copilot, or ChatGPT Team, or a workflow automation platform, point it at the team, and watch time come back.

It doesn't work like that, and it's not the software's fault.

AI tools are genuinely good at specific things: summarising documents, drafting routine communications, pulling patterns out of data, flagging things that need attention. But they work on whatever process they're plugged into. If that process is patchy, inconsistent, or built on ten years of workarounds and institutional habits nobody has examined, the AI will accelerate the mess rather than clean it up.

The DSIT research also found that the skills gap is the primary barrier for over 60% of UK businesses trying to adopt AI. But "skills gap" is often shorthand for something more specific: people don't know where to apply AI because nobody has mapped out where the real friction sits. You can't automate your way out of a workflow you've never properly documented.

What actually works

The businesses that do see returns from AI share one thing in common. They looked at their workflows before they looked at tools.

That means sitting down with the people who actually do the work, not just the managers above them, and mapping out how a task moves from start to finish. Where does it slow down? Where do things have to be manually moved from one system to another? Where does someone spend twenty minutes each day doing something a computer could do in five seconds? Those are the spots where automation genuinely earns its place.

Microsoft's own data suggests UK managers spend around 15.5 hours every week managing emails and meetings. That's nearly two full working days. Co-pilot can help with meeting summaries, email threading, and draft responses, but only if it's deployed alongside a clear understanding of how email and meetings are actually used in your organisation. A summary of a meeting nobody needed to attend saves nothing. An action list from a well-run weekly catch-up saves hours.

The businesses getting real value from AI aren't doing something dramatically different from everyone else. They've just done the groundwork first: mapped the workflow, identified the friction, chosen the right tool for the right job, and trained their people on something specific rather than hosting a one-off webinar and hoping it sticks.

What to check if your AI tools aren't delivering

If your business has AI tools running but the team isn't feeling the difference, it's worth working through a few honest questions.

1. Do you know where your biggest time drains actually are? Not where you think they are, but where your team genuinely loses hours each week. Manual data entry, approval chains that require chasing, documents saved in the wrong place and searched for in meetings; these are the usual suspects. If you haven't mapped that out, start there.

2. Has anyone shown the team how to use it in the context of their actual job? Generic onboarding doesn't land. A conveyancer needs to know how Co-pilot helps with document review, not how it summarises Wikipedia pages. Training that's specific to the role is what drives adoption.

3. Are your core systems connected? AI tools perform best when the underlying data is clean and joined up. If your CRM, your project management tool, and your finance system don't talk to each other, the AI is working with fragments.

4. Is this one tool solving three things, or three tools solving one thing each? Licence sprawl is a real problem in SMEs right now. Auditing what you have, what's actually being used, and what overlaps can often surface both savings and better results from what you already own.

5. Have you given it enough time with proper support? Six months with a webinar is not an AI adoption programme. Real change takes closer to twelve months with regular check-ins, clear use cases, and someone accountable for making it work.

A note on Microsoft 365 pricing

One practical prompt to act now: Microsoft is increasing its 365 pricing from July 2026, with changes that reflect the broader integration of Co-pilot across its products. If your business is approaching a renewal or hasn't reviewed its M365 licensing recently, this is a good moment to understand exactly what you're paying for and whether your team is getting full value from it. The Microsoft 365 Roadmap gives a clear picture of what's coming across the product suite.

Getting it right

AI is not a shortcut. It's a multiplier. Which means if your workflows are working well, the right AI tools can make your team significantly more productive. If your workflows have problems, AI will make those problems faster and more consistent.

The good news is that most workflow problems in SMEs are solvable, and they don't require expensive consultancies or full digital transformations. They require someone to sit down with the team, ask the right questions, and build a clear picture of where the friction is.

That's exactly the work our team at BITS Group does as part of our workflow optimisation service. If you're paying for AI tools and not seeing the return you expected, it's probably worth having that conversation. Drop us a line at bitsgroup.co.uk and we can take it from there.

Further reading

• DSIT: AI Activity in UK Businesses (UK Government)

• Microsoft 365 Co-pilot: What's New, March 2026 (Microsoft Tech Community)

• Microsoft 365 Roadmap (Microsoft)