What artificial intelligence can really do in a small business
AI does not replace your trade. It takes on the tasks you do badly because they repeat: following up, recording, reposting, remembering. Rarely the ones people think of handing over first.
Published in September 2026
The right question is not which AI, but which task
Most owners approach the subject through the tool. The better way in is elsewhere: list what you redo identically every week. If a task can be described in one sentence and repeats, it is a candidate.
The test is three questions. Do I do this more than twice a week? Does the result always take the same shape? Does a mistake here cost little? Three yeses, and the task can be handed over.
What works, and what we commonly install
These five uses show a visible effect in under a month.
- Follow-up: a quote with no reply is chased at three days, then at ten, without anyone thinking about it. Nine unanswered quotes out of ten are an oversight, not a refusal.
- Routine writing: quotes, reports, answers to frequent questions, written in your vocabulary and reviewed by you.
- Publishing: the week's texts and visuals, prepared in advance, ready to post.
- Case memory: what was said to whom, and when, without depending on a notebook or a phone.
- Sorting incoming enquiries: separating what deserves a call from what deserves a standard reply.
What does not automate, and should not be attempted
Negotiating a price. Visiting a site. Deciding to trust a supplier. The moment an unhappy client needs to hear a voice. These call for judgement, and judgement handed to a machine turns against you sooner or later.
Our rule, with our clients: the machine opens the door, a human takes over the moment the conversation matters. And on prices, always a human.
Where to start, concretely
Take the task that irritates you most in an ordinary week, not the most strategic one. That is the one that will let you see the effect, and a visible effect is what decides the rest.
Allow two to three weeks for a first automation to be installed, tuned and judged on results. If you see no difference by then, it was aimed at the wrong task.
Questions that come up
Do you need a large budget to start?
No. A first useful automation covers a single task and is set up in a few days. What costs money is stacking up tools bought without a precise use.
Does my data leave the company?
That depends entirely on the setup. Ask where your data lives, who has access, and what happens if you stop. A provider who cannot answer those three questions clearly is the wrong one.
Will it replace my employees?
In the cases we see, no: it removes the repetitive work and gives them back time in the field. A company that automates in order to cut staff ends up with automated tasks and nobody left to judge the result.
Is your case different?
Tell us where you stand. You get an answer about your situation, not a sales pitch.