Choosing a first AI project: start where a department is willing to own it

What makes a good first project

• An owner: a named person in the department who wants the result and can give real time to it.
• One clear task: work that is done often, such as answering the same questions or copying details from forms into a system.
• Data you already have: information the team holds today, is allowed to use, and that IT has agreed to.
• A way to tell if it worked: a measure agreed before the project starts, such as hours saved or questions answered.

Try a tool or get outside help?

• A ready-made tool fits a common task for one team that does not need to connect to your own systems.
• Outside help makes sense for a task that connects to your own systems or data, or needs work built around it.
• Both: some organisations try a tool first to learn what they need, then get help with the parts the tool cannot do.

Plan what happens if it works

Before you start, decide who will run the system after the pilot and where the budget for the next stage will come from. A pilot that works but has no next step can stall as easily as one that fails.

Frequently asked questions

Should we pick the idea with the biggest savings?
Not always. A smaller idea with a keen owner and data that is ready can show results sooner, and that can help the bigger ideas get support later.

Do we need a full AI strategy first?
Not necessarily. A short review of a few ideas against the four points above can be enough to choose a first project.

What is a good first step?
List your AI ideas. For each one, write down who would own it, what data it needs, and how you would know it worked. Start with the idea that has the clearest answers.

---

Choosing your first AI project? Talk to our team.