The accusation does not hold up
The story goes that smaller firms are slow, cautious and losing ground. The numbers say otherwise.
- A quarter of UK businesses now use AI in some form
- More than half of UK firms are actively using it
- Nearly two-thirds of UK directors use AI personally
What is genuinely true is that adoption is shallow. The figures measure whether AI is present. They do not measure whether it has changed how the business runs. That is the real gap, and it is a very different problem from resistance.
Two risks, and you cannot dodge both
Underneath most hesitation sit two risks.
Risk A is uncontrolled adoption. Pilots without ownership. Tools without rules. Outputs trusted without checking. Customer data pasted into free consumer accounts. Errors that ripple into quotes and contracts.
Risk B is uncontrolled avoidance. Delay dressed up as caution. Staff using AI anyway, with no policy and no training. Skills stagnating while competitors compound their gains. Good people leaving for firms that have built the capability.
A "no AI" position does not remove either. It usually just removes leadership from the conversation.
The risk nobody chose
The most overlooked risk is shadow AI - staff using AI tools outside any policy.
Around 71% of UK employees have used unapproved AI tools at work, and over half do so weekly. Yet only about a third of organisations have a formal AI policy. Research into data breaches found firms with high levels of shadow AI carried average breach costs several hundred thousand pounds higher than those without.
For a small business, this is not abstract. It is a customer's confidential enquiry pasted into a free account to speed up a quote. It is board figures drafted in a personal AI tool.
Banning it does not help. A ban pushes the same behaviour further out of sight. Bringing it into the open is what actually reduces the risk.
Why so many pilots go nowhere
Business owners who have watched an AI pilot stall are not being irrational. Research on AI projects consistently finds the same causes of failure.
- The business problem was misunderstood
- The data was not good enough
- AI was applied to the wrong problem
- Nobody owned the workflow
- No success measure was agreed before starting
Very rarely is the technology itself at fault. It is the conditions around it.
There is an engineering parallel worth holding onto. Jumping to autonomous AI agents before mastering the basics is like switching on someone else's machine to run unattended without having built it, commissioned it or read the manual. You build it. You understand the components. You run it manually. Only then do you let it run alone.
Where a small business should actually start
Not with the most complex, highest-stakes part of the operation. Start where the data is usable, the risk is bounded and the outcome is measurable.
- Document summarising and knowledge retrieval
- Quote drafting and enquiry response
- CRM enrichment and prospect research
- Management reporting and analysis
- Customer-service triage with human escalation
- Email management and proposal drafting
For most businesses that means the office before anything operational. It is also where the owner spends a meaningful share of their own week, which means they can model the behaviour rather than delegate it.
What controlled adoption looks like
The pattern is consistent. A named senior owner. A plain-English acceptable-use policy. Pilots tied to real workflows rather than "let's see what AI can do." A success metric agreed before the pilot starts, so it can be killed or scaled on evidence. Human review designed in rather than bolted on. Training funded alongside the licences.
None of that is exotic. It is the same discipline any well-run business already applies to a new process or a piece of capital equipment.
The full article, by Sales and Marketing Engineers covers the UK research behind each of these points, why pilot fatigue is a rational response, the CONTROL framework for anchoring AI decisions, and a fuller breakdown by role - owner, commercial and operations. The statistics and full citations all sit in the main article - read more here
Author
Stefan Buss BSc Eng (Ind) , Founder of Sales and Marketing Engineers Ltd, has been a member of gdb, in different forms, for over 12 years.
