What actually happens in an AI consultation
An introduction came with a warning attached. Very hands on, this one. Very reactive. Drops everything the moment anything happens in the business.
That turned out to be completely accurate, and it also turned out to be the diagnosis.
The owner should have been adjusting things. Instead he was doing the heavy lifting, all day, every day, and every new thing that arrived went to the top of the pile because he was the only one who could deal with it.
No AI project fixes that on its own. But you cannot have a useful conversation about AI in a business like that without starting there, because the constraint is not the technology. It is where his attention goes.
Most of the first hour is not about AI
People are sometimes surprised by that.
A useful first conversation is mostly: what happens in a week here. Who does what. Which bits get dropped when it is busy. What you would do with a spare afternoon if you had one.
Then, usually near the end, a few obvious candidates fall out. Not because I brought them. Because saying out loud what actually happens in a week makes the repetitive bits audible in a way they are not when you are inside them.
That is why the diagnostic is 45 minutes rather than a proposal document. You cannot cost anything sensibly until you know where the time goes.
Where most UK businesses actually are
It is worth knowing the real numbers, because the coverage will have given you the impression you are behind.
The Department for Science, Innovation and Technology found 16 per cent of UK businesses with five or more employees were using AI in spring 2025, with 80 per cent neither using it nor planning to. Their fieldwork is older than some other surveys and telephone research probably misses informal use, which they say themselves.
The ONS, asking more recently, found 28 per cent of micro businesses using at least one AI technology, against 49 per cent of businesses with 250 or more staff.
So yes, bigger firms are ahead. But the gap is 28 against 49, not 5 against 90. Most of the country is somewhere near the start of this, including plenty of businesses that look like they are not.
You are not as far behind as the LinkedIn posts suggest.
What a consultant should not do
Turn up with a strategy.
An AI strategy for a fourteen person business is usually a document that makes everyone feel organised and changes nothing. What that business needs is one thing that works, finished inside a fortnight, so the people in it stop treating AI as a thing that happens to other companies.
The second one is easier. The third one they find themselves.
What you could do on Monday
Before you speak to anybody, including me, do this.
Write down the last five things that landed on your desk that should not have needed you. Not the important decisions. The ones where you were just the person who happened to be able to do it.
That list is the shape of your business's actual problem. Some of it is delegation to a person. Some of it is a process nobody wrote down. And some of it, usually two or three items, is genuinely automatable.
Any decent consultation starts from something like that list. If you arrive with it already written, you will get a great deal more out of the conversation.
The honest limit
Some businesses should not be doing this yet.
If you are in the middle of moving premises, or changing your accounting system, or you have just lost two people, the answer is almost always not now. AI work needs a bit of slack to land in, and a business with no slack will start it, get interrupted, and conclude it does not work.
There is also a category where the honest answer is that the problem is not one AI solves. A business with no pipeline does not need a faster quoting process. It needs customers, and I would rather say so in the first half hour than sell you a project that goes in the wrong direction very efficiently.