Five questions to answer before your first AI project

Almost every leadership team we speak to wants to "do something with AI". Far fewer have decided what that something is. Before choosing a tool or a vendor, work through these five questions.

1. Which problem are we solving?

Start with a task that is frequent, time-consuming and well understood: summarising reports, answering routine customer questions, extracting data from documents. If you can't describe the task clearly to a new employee, an AI system won't do it well either.

2. What does good look like?

Decide how you'll judge the result before you start. Time saved per week? Fewer errors? Faster response to customers? Without a measure, a pilot can't succeed or fail. It just continues.

3. Is our information ready?

AI tools work with the information you give them. If policies are out of date, records are inconsistent or key knowledge lives only in people's heads, fix that first. This work pays off even if you never deploy AI.

4. Who checks the output?

AI makes mistakes, sometimes confidently. For anything that affects customers, money or compliance, decide who reviews the output and how. Keep a person accountable for every decision the AI informs.

5. What are our rules?

Agree simple guidelines before staff start using AI tools on their own:

• What information can and can't be shared with external AI services
• Which tasks AI may help with, and which it may not
• How AI-assisted work is labelled and reviewed

These rules protect your customers and your staff, and they make adoption faster because people know where the lines are.

Start small, then scale

Pick one well-defined task, run a short pilot with clear measures, and review honestly. A successful small project builds the skills, data habits and confidence you'll need for larger ones.

---

Our Advisory team runs AI readiness assessments and helps organisations choose their first use cases. Start a conversation.