Most leadership teams we talk to are not short of AI ideas. They are short of a way to choose between them. A list of twenty possibilities — a chatbot, forecasting, automated reporting, a smarter CRM — usually ends in one of two places: nothing happens, or the most visible idea wins and the business learns very little from it.
The first initiative matters more than its size suggests. It decides whether people trust the next one, it shows what your data and systems can really support, and it sets the habits — ownership, measurement, review — that later projects inherit. So it is worth choosing deliberately.
Start from a decision or a task, not from the technology
“Where can we use AI?” is the wrong first question. A better one is: which recurring decisions or tasks cost us the most time, money or missed opportunity?
Write them down in plain business language:
- “We reorder stock from gut feel and run out of our best sellers.”
- “Our team spends two days every month assembling the same performance report.”
- “Customers wait hours for answers that are already in our policies.”
Each of these has an owner, a frequency and a cost. That makes it something you can improve and measure. “Implement AI in operations” has none of those things.
Score each option on four questions
For every candidate on the list, ask:
- How much is it worth? Estimate the value in hours saved, revenue protected or errors avoided. A rough range is enough; the point is to compare options with each other.
- Is the data there? The information the task depends on must exist, be reasonably complete and be reachable. If the answer is “it’s in someone’s head” or “it’s spread across five spreadsheets”, the first project may really be a data project.
- How contained is it? A good first initiative touches one team and one or two systems. Anything that needs every department to change at once is a second-year project.
- What happens when it is wrong? AI output is sometimes wrong. Prefer work where a person reviews the result before it matters — a draft, a recommendation, a flagged exception — over work where a mistake goes straight to a customer or into the books.
The strongest first candidates score well on all four: meaningful value, available data, a contained scope and a low cost of error.
Be wary of three common traps
The showcase project. A customer-facing assistant or an “AI strategy for the whole company” gets attention, but it is hard to scope and harder to measure. Visibility is not the same as value.
The tool looking for a problem. Buying a platform first and finding a use for it later reverses the logic. The business need should choose the tool.
The pilot with no finish line. If you cannot say in advance what result would justify continuing — or stopping — the pilot will drift. Agree the success measure before any work starts.
Define what “done” looks like
Before committing, write a one-page brief for the chosen initiative:
- the task or decision it improves, and who owns it;
- how it is done today, and roughly what that costs;
- the data and systems involved;
- what a person still reviews or approves;
- the measure of success, and when it will be checked;
- what would make you stop.
If that page is hard to write, the initiative is not ready yet — and you have learned that cheaply.
Small, real and measured beats large and theoretical
A first initiative that saves one team a few hours a week, runs reliably and is trusted by the people who use it is a success. It gives you evidence, a working pattern and the confidence to take on something larger.
That is the approach we take with clients in our Advisory work: narrow the options to the few that matter, check what the business can support, and leave you with a first step you can act on. If you are weighing your own list of ideas, tell us about it.