Lean has a branding problem
Lean has a branding problem. Mention it to most business owners and they picture a manufacturing floor, a whiteboard covered in sticky notes, or a consultant with a clipboard counting how many steps someone takes to fetch a part. That is a shame, because the actual idea underneath lean is simple and still correct: find the waste, remove it, and do this constantly rather than once a year.
The reason lean has stalled in a lot of small businesses is not the thinking. It is the overhead. Proper lean work takes time to observe a process, time to document it, and time to test a change. Owner-operators running a £2m business do not have a spare afternoon a week for that, so the sticky notes go up once, during a training day, and come down a month later when nobody has kept them current.
This is where AI is quietly useful, and not in the way it gets sold.
The waste was always in the noticing
Most of the friction inside a small business is not dramatic. It is a manager re-typing the same numbers into two systems. It is an admin person reading forty emails a day to find the six that need a reply today. It is a founder checking stock levels manually because nobody trusts the spreadsheet to be current. None of this looks like waste on a org chart. It only looks like waste when you watch someone do it.
That is the actual lean skill: noticing. And it turns out a well-built AI tool is quite good at the boring, repetitive noticing that a human gets tired of doing. It can watch for the same pattern every day without losing attention on day fourteen. It can flag the six emails that matter. It can reconcile the two systems automatically instead of relying on someone remembering to.
Put simply: lean asks “where does the process leak,” and a well-scoped AI tool is often the cheapest way to actually see the leak, rather than guessing at it from a workshop.
Small, and in the right order
The lean principle that matters most here is not the tools, it is the sequencing. Fix the biggest recurring drag first. Fix one thing at a time. Prove it works before building the next thing. This maps almost exactly onto how we think AI should be introduced into a business, and for the same reason it works in lean: a business absorbs one well-built change far better than it absorbs five half-built ones at once.
The businesses that get this wrong tend to buy a platform first and ask what problem it solves second. That is the opposite of lean thinking, and it is also, in our experience, the most common way small businesses end up with an unused AI subscription and a bad taste about the whole category.
Where this actually lands
Take a hypothetical wine merchant with a stock reconciliation problem: three hours a week of one person manually cross-checking supplier invoices against what the till said had sold. Not a big number in isolation. Multiplied over a year, and over the cost of the person doing it instead of something else, it is worth building around. The fix would not need to be clever. A small tool that reads the invoices and flags mismatches turns those three hours into twenty minutes of checking exceptions.
That is what lean looks like done with AI rather than a whiteboard. The audit still matters. The walking of the floor still matters. What has changed is that the fix, once you have found the friction, is often faster and cheaper to build than it would have been three years ago.
The part worth remembering
None of this replaces the people already doing the work. The point is not a leaner headcount. It is a business that spends less of its week on the parts that were never the point, and more of it on the parts that were.
If lean is a discipline of noticing and removing waste, AI is not a replacement for that discipline. It is, for the first time, a genuinely cheap way to practise it.