Skip to content

The right first AI project is narrow, measurable and close to real work. A short guide to picking one that earns its keep.

AI & Data Science1 min read

Most organisations do not need an AI strategy before they need one useful AI tool. The quickest way to learn what AI can do for your business is to put a well-chosen project into the hands of the people who will use it.

Narrow beats ambitious

A first project should solve one problem for one team. Summarising incoming documents, drafting replies to common enquiries or flagging unusual transactions are all small enough to build quickly and clear enough to judge.

Check your data first

AI is only as good as the information it can reach. Before you commit, confirm the data exists, that you are allowed to use it for this purpose, and that someone owns its quality.

  • Where does the data live, and who can grant access?
  • Does it contain personal or sensitive information?
  • How will you know if the answers are wrong?

Keep a person in the loop

Early projects work best as assistants rather than decision makers. Let the tool prepare the work and a person approve it. You keep control, your team builds trust in the tool, and you collect the feedback you need to improve it.

All insights