Here is how it usually happens. You used ChatGPT to rewrite a difficult email and it came out better than yours. You used it to summarise a forty-page document before a meeting. A colleague noticed. Then, in a leadership meeting you were not in, someone said “we should be doing something about AI”, someone else said your name, and now you are the AI person.
You did not sign up for this. You are also, almost certainly, the right person for it, and this page is about why, and what to do in the first thirty days.
What the boss is actually asking
Nobody wants a strategy document. When an owner says “what should we do about AI”, they are asking three plainer questions, usually without knowing it:
- Where would this save us real money or time? Not “transform the business”. Hours a week, on a named task.
- What could go wrong? A customer’s data pasted somewhere it should not be. A wrong figure nobody caught. A tool somebody built that nobody else understands.
- Who is going to own it? They are hoping the answer is you, and hoping they do not have to hire someone.
You can answer all three in a month, and none of the answers needs you to understand how the technology works. They need you to understand how the business works and to be honest about what the tools can do. That is the whole job description, and it is why a vendor, who has something to sell, cannot do it for you.
The three levels of “using AI”
Before you answer anything, get one distinction straight, because it is the one the whole company is confused about. “Using AI” means one of three things.
Level 1: people use it. Rewriting, summarising, asking questions. This is where you started and where nearly everyone is. It makes individuals faster at tasks that were rarely the bottleneck. It is worth having, and it is not a strategy.
Level 2: a process changed. Not a person being faster, but a hand-off disappearing. Quotes that go out the same day. Enquiries sorted before a human reads them. This is where the money is, it rarely needs new software, and it is what the boss is actually hoping for when they say “AI”.
Level 3: software does the work. A tool built for the business, possibly by you, possibly by someone in accounts who watched a video. Genuinely powerful. Also the single biggest new risk in a small company, because the tools make the app easy and do nothing for everything around it.
When you report back, say which level you are talking about. It will make you the clearest person in the room.
Your first thirty days
Week one: find out what is already happening. Ask people whether they use AI and most will say no. Ask whether they have pasted an email thread or a customer list into one and most will say yes. Your company already has an AI plan; it is just unwritten and different for every person. Make a list of which tools, on which data, by whom. Do not scold anyone. You are collecting, not policing.
Week two: write the one-page rule and walk people through it. Which named tools are allowed. What may go in, what needs care, what never goes anywhere. Who to ask, which is you. There is a template; fill in the names and it is done. Then brief it in person, on the company’s own documents, because a policy email is filed and never read.
Week three: walk the business, hand-off by hand-off. Sit with each team for an hour and watch how work moves: where it is re-keyed, chased, waited for, checked. Write down every candidate for level two. Score each on payoff and effort. The method is two weeks; you are doing the fast version.
Week four: pick one, and say “not yet” to the rest. Take the top candidate and change that one process, with the people who do it. Measure the hours before and after. That single number, on one process, is worth more to the boss than any slide deck, and it earns you the right to do the next one.
What to say no to, for now
- “Can we get a chatbot on the website?” Not until you know what it would answer from and who checks what it says.
- “Sarah built a tool over the weekend and sales love it.” Wonderful. Before the business relies on it, someone other than Sarah checks it, it runs under a company login, and there is a backup someone has actually restored. The checklist is short. Your job is to require it, not to do it.
- “A vendor says their AI will transform our business.” Ask them these questions and watch which ones go quiet.
- “Should we hire an AI consultant?” Not before weeks one to three. After them, you will know exactly what to ask one for.
The thing that keeps you in the role
The fastest way to lose the AI person job is a bad week: a customer’s data in the wrong place, a tool that gave a wrong figure for a month, the one person who understood the weekend app leaving. None of these need you to become an engineer. They need someone to have asked the questions in advance, and the questions are written down for your boss.
The slowest way to lose it is to let the company learn the same lessons about AI over and over. Keep a page of what worked, what did not, and why. Written once, read by the next person. It is the part that compounds.
If you also write code
Some AI people are also the person who now builds things. That is a different page, with a different set of questions your company will ask of you: The informed way to code with AI inside a company.
If you want a second opinion
I work with a small number of clients, and some of them are people who have just been handed this role, because I have been that person: I came to software from the business side, and became the one people asked long before I could build anything. Say hello and tell me what week you are in.