Is AI worth it for a small business? An honest answer.
Most of the small business owners I talk to in 2026 think AI is overhyped. They are not entirely wrong, and the question “is AI worth it for a small business” deserves a more honest answer than the one it usually gets.
In the run-up to launching Cordial Advisory I have had this conversation, in some form, with maybe twenty or thirty people. Founders I know from the design and luxury sectors, contacts from my time in travel, old colleagues, friends of friends. The pattern is consistent. They have tried ChatGPT. They have signed up to two or three AI tools that turned out to do almost nothing useful. They have read the headlines. And they have arrived, quietly, at the view that AI is mostly noise.
I will say it plainly. A lot of what is sold to small businesses as AI is not worth their time. It is repackaged software with a chatbot bolted on the front. It is a subscription that solves a problem they did not have. It is a long onboarding for a small return. The cynicism is earned.
But the conclusion most owners have drawn from that, that AI as a whole is not for them, is too neat, and it is costing them. There are two tiers of useful here, and most owners are missing both. The first is everyday. The second is built. They work differently and are worth treating separately.
The everyday tier
This is the bit that gets undersold by people like me, because it is unglamorous and we do not get paid for it. But it is where the largest number of small business owners can get real value this week, with no spend and no consultant in sight.
ChatGPT, Claude, and the other general-purpose tools are, if you use them properly, the most useful thinking partner most small business owners have ever had access to. Not for writing your customer-facing copy, where they are mediocre at best. For the work in between. Drafting the difficult email to the supplier. Sense-checking a contract before you send it to your solicitor. Summarising a forty-minute meeting recording into the three decisions you actually need to act on. Turning a messy brain-dump into a clean one-pager for the team. Comparing two pricing structures side by side without spending a Sunday on it.
I use one of these tools for somewhere between an hour and three hours a day, across roughly fifteen tasks of that shape. None of them are individually impressive. None of them are the kind of thing I would write a case study about. But together they are the equivalent of having a thoughtful, fast, slightly junior colleague available at any hour, and they cost me twenty pounds a month.
For a small business owner who has not done this properly, the single highest-return move you can make this year is to spend a fortnight using one of these tools, every day, on the work you actually do. Not asking it to “write a marketing strategy.” Giving it the messy email you were about to send and asking it to tighten it. Pasting your last board pack and asking what is missing. Walking it through a hiring decision and asking what you have not considered. That is where the return lives.
Most owners I talk to have not done this. They have tried it three times, asked it to do something it is bad at (like writing finished copy in their voice), been unimpressed, and concluded that AI is not for them. They are judging the technology by its worst use case. It is like deciding cars are useless because you drove one in first gear.
The built tier
Then there is the second tier, which is what the consultancy work and the AI-tool companies are mostly selling. Custom tools, built around a specific job in your business, that take a repetitive task and shrink it. This is where the dramatic time savings come from, but it is also where most of the hype lives, and where most of the wasted money goes.
AI works here when the job is repetitive in shape, structured enough to teach, and important enough to senior people that getting forty per cent of it back genuinely changes the week. That is a narrow set of jobs. But when one of them is hiding in your business, the return is real.
An example, because abstract is not useful. It is from my last role, before I started this, so I can speak to it properly.
I spent four years as Managing Director of a luxury sculpture studio, around thirty-five people. The quoting process took the senior salespeople a real chunk of their week. They pulled prices from supplier lists, applied a margin rule that varied by product type, wrote the quote into a document, and emailed it across. Each one took the better part of an hour. Across the team, it was most of a day a week, gone to the same job, every week, done by people whose time was worth more on the phone with clients.
We built a tool that took the inputs from a short form, applied the margin rules, pulled the supplier prices automatically, and produced the same document the team used to write by hand. It saved about forty-five minutes per quote. Across the salespeople it added up to roughly 2.5 hours a week each. They still read every quote before it went, because they should. But the heavy lifting was no longer theirs.
That sounds modest until you do the maths. Two and a half hours a week, across four salespeople, is ten hours a week. Over a year, that is roughly thirteen working weeks given back to the team. And those thirteen weeks went into client conversations, not paperwork, which is why the same period saw a twenty per cent uplift in business development opportunities. The tool did not create that number on its own, but it bought the time that allowed it to happen.
That is the kind of return the built tier gets you. Not transformation. Not a different business. A real shift in what the senior people in your business actually spend their week doing, and a return on a one-off build cost that pays for itself in the first three months and keeps paying after that.
What it is not, importantly, is everywhere. The same business had plenty of other jobs that looked automatable on paper and were not. Anything that involved judgement about a client relationship. Anything that required reading between the lines of an email. Anything where the rules were real but unwritten and lived in the head of the person who had been doing the job for fifteen years. We looked at those and left them alone, because the tool would have done a worse job than the person, and the saving would have been bought at the cost of a quieter, slower problem down the line.
A test for the built tier
The everyday tier you can just try. There is no decision to make beyond opening the tool and using it on real work for a fortnight.
The built tier needs a test, because the cost of getting it wrong is higher.
Take a job in your business that someone senior does every week. Ask three questions about it. First, is the job repetitive in shape, even if the content changes each time. Second, is the input to the job structured, or could it be made structured with a short form. Third, is the output something the person could check in a minute, rather than something they would have to redo from scratch if the tool got it wrong.
If the answer to all three is yes, the job is probably worth automating, and the return will be real. If the answer to any of them is no, leave it alone, regardless of what the AI tool’s website tells you.
Most jobs in a small business will fail at least one of those questions. That is not a failure of AI. It is the honest shape of where the technology is in 2026.
The owners who get value out of AI right now are not the ones using the most of it. They are the ones who have done two things. They have made the everyday tier part of their working week, properly, so that the small returns compound. And they have found the two or three jobs in their business that pass all three questions of the built tier, and left the rest of the business alone.
That is the whole game, more or less.
If you want a second pair of eyes on which two or three jobs in your business pass that test, that is most of what an audit is. Either way, I write something like this most weeks. You are welcome to follow along.