I ask the same question on almost every first call. Who on your team already uses AI, and what have you told them they can and cannot do with it. Most owners pause. Then they say something like, I think Maria uses it for emails, or, I'm not sure. The owner in that pause isn't careless. Nobody on the team sat down and decided who could use what. A tool showed up on someone's laptop, it worked, word spread through the team, and a few months later half the staff use it while the other half don't know it exists. Nobody chose that outcome. It grew on its own, unsupervised. When nobody sets the rules, the first mistake sets them instead.
This pattern shows up in businesses with real structure, not only in small operations running loose. Scale the survey up to a team of 10 to 50 and the number barely moves. The tools spread faster than the conversation about them. The business two doors down might already have a page pinned to the staff room wall. They aren't more advanced. Someone there gave it twenty minutes on a slow Tuesday, and that was the whole difference. A separate 2026 survey found that 58 percent of employees have pasted sensitive information into an AI tool without asking anyone first, client records, internal numbers, the kind of detail a customer expects to stay private. None of them meant harm. Typing it in was faster, and nobody had told them where the line sat. Wait for the perfect moment to fix this, and a customer notices first: two staff members give two different answers to the same question, one using AI and one not. By then the fix costs more than twenty minutes. Most of that gap comes down to five decisions nobody has made yet1 / Customer data walks out on someone's laptopEvery AI tool sits between your team and whatever they type into it. Nobody decided what's allowed to cross that line, so the decision sits with whoever is busiest that day. A rushed reply with a customer's full history pasted into it can leave your business before anyone reviews it once. 2 / One missed check costs more than the time it savedSome AI drafts need a second pair of eyes before they reach a customer. Others don't. Businesses that skip this decision land on one extreme: checking everything and burning out the person doing the checking, or checking nothing and letting an error reach a customer who complains about it in public. 3 / When something breaks, everyone points at someone elseAn AI mistake with no owner turns into an argument instead of a fix. If your team can't name who handles it inside five seconds of you asking, that gap matters more than anything else on this list. 4 / Five staff, five different standardsAsk five staff what a good AI-written message looks like, and expect five different answers. None of them agree, and customers notice the inconsistency before you do. One real estate office I worked with had five agents describing a finished listing five different ways. One page fixed it in a day. 5 / Your team already contradicts itselfAsk your front desk person and your driver the same question about AI, and expect two different answers. Customers ask the same question to different people all the time, and they compare notes more often than most owners assume.
This year alone, three clients walked into exactly this situation with no warning: a hotel, a dental practice, and a transport company. Each one had staff already using AI on their own for weeks before we spoke. None of the three owners had written down a single rule. In each case, the five decisions above took under an hour to turn into one page, pinned somewhere the team looks every day. The hotel used it to settle which guest messages needed the manager's eyes first. The dental practice used it to draw the line on patient data in plain words instead of legal language nobody reads. The transport company used it to stop two different answers reaching the same customer on the same day. At the hotel, a staff member almost sent a guest a discount code the hotel had never offered. The AI draft stated it as fact, and nobody checked before it nearly went out. Someone happened to reread the message in time. Next time, they might not. All three owners went from being the last to know what their staff were doing with AI, to having one page that outworked a full policy, because their teams read it.
Most advice about AI tells you which tool to buy next. Decide first what your team can do with the tools they already have. Every business I've worked with that skipped this step paid for it later, in a mistake that was small on its own and expensive only because nobody had agreed what should have happened instead. It is the cheaper problem to solve. Most businesses skip it. Talk soon, Mikkel P.S. Reply and tell me what tool your team already uses without asking anyone. I read every reply, and it tells me exactly which section of the House Rules your business needs first.
The AI Solopreneur • AI Strategy for European SMBs |
