The UK economy’s 0.1% stagnation has exposed a widening AI friction tax, where high AI adoption rates are being undercut by a messy transition period that’s stalling national productivity.
Despite over a third of SMEs integrating AI into their daily work, many are finding that a fumble period – the gap between buying a tool and learning to use it effectively – is unintentionally creating more work than it saves.
While falling business investment and weak consumer spending have also dampened the UK economy’s momentum, remedying the disconnect between productivity and adoption has never been more important. To understand how, it’s worth looking at similar challenges in the recent past.
Why hasn’t AI improved UK SME productivity yet?
History suggests this period of friction is a standard part of the journey toward becoming more efficient. Economists call it the “Solow Paradox” – the idea that you can see new technology everywhere except in growth figures. In the early 2000s, businesses didn’t get faster just because they bought computers; they got faster when they stopped using them as “fancy typewriters” and redesigned their entire workflow.
Right now, many SMEs are in that same transition. We’re fitting lightning-fast AI into slow, manual processes. Research from Starling Bank suggests this lack of integration saps £25.3 billion from the economy annually, as staff spend 15 hours a week manually bridging the gap between disconnected apps. It’s not necessarily a failure of AI, but a sign that the “fumble” is a real, measurable drag on our time.
What are the signs of AI friction tax in the workplace?
Prompt-Looping:
Staff spend 20 minutes or more trying to get a chatbot to write a two-minute email. It feels like work, but it’s a technical hurdle that hasn’t actually saved any time.
The Accuracy Audit:
Managers spend more time fact-checking and editing AI-generated drafts than they would have spent writing from scratch. If a senior lead spends an hour fixing a five-minute AI output, the gain is lost.
The Over-Polish:
Using the time saved by AI to endlessly tweak internal documents that would’ve previously been good enough. This creates a volume of beautifully formatted content that doesn’t actually help ship a product or close a deal.
How can UK SMEs overcome the AI learning curve?
Russell Dias, Product Manager for AI at Employment Hero, suggests psychological barriers may be one of the main drivers for teething issues with AI adoption.
“I think adoption stalls when the expected downside – being wrong, non-compliant, or unsure how to start – feels bigger than the time saved.”
Despite this, he points out that “fear of hallucinations is very overblown”.
Adding: “It’s definitely not a solved problem, but it’s simultaneously getting better and harder to detect as models get smarter. A lot of people struggle to understand what good usage looks like. In tech or coding, it’s becoming clear what really good looks like, but for most other roles this is very difficult to tell.”
Three Steps to Overcome the AI Learning Curve:
Russell suggests that to exit the fumble period, businesses need to move away from treating AI as a vending machine of sorts and start treating it as a workflow, starting by:
1. Making the rules crystal clear
Set clear guidelines on what tools are acceptable, what data should never be input into these tools, and what good review looks like before sending. There’s a lot of noise around privacy, IP, and compliance – cutting through it helps employees move forward with confidence.
2. Finding and empowering your power users
Identify employees who are already using AI effectively and have them share examples through show-and-tells. This isn’t about big training courses – it’s about endorsing usage, normalising it, and showing what’s possible in real workflows.
3. Starting with no-brainer use cases
Begin with simple, high-impact tasks: first drafts (emails, proposals, job ads), summarising long documents or meeting notes, and simple analysis or explanations from non-sensitive data. Build confidence before tackling more complex workflows.
Will agentic AI solve the productivity paradox in 2026?
The shift in 2026 is away from simply chatting with AI and toward deploying it. UK workplaces are moving toward Agentic AI – autonomous agents that live inside your payroll, HR, or supply chain software to execute work without you having to ask every time.
By embedding AI into the daily flow of work, businesses can bypass the need for constant human mediation. It’s one way to stop talent from being wasted on auditing AI outputs and start using it to drive the growth that’s currently missing from the national stats. We’ve got the tools; now we have to stop fumbling with them and let them run.
