Choosing the first AI project is the hardest step for most teams. Pick something too big and it stalls. Pick something too small and nobody notices. This guide shows how we help clients choose a first use case that delivers clear results.
List every repetitive task
Ask each team to list tasks they repeat every week. Include how long each one takes and how often mistakes happen. You will usually end up with 20 to 40 ideas.
Score each idea
Give every idea a score from 1 to 5 on four simple questions:
- Value: how many hours or how much revenue does it affect?
- Data: do we already have clean examples to learn from?
- Risk: what happens if the AI gets it wrong?
- Speed: can we ship a working version in four weeks?
Add the scores and keep the top three. These are your candidates.

Run a small pilot
Pick the highest scoring idea and run it with one team for four weeks. Keep a person in the loop to review every result, and track the time saved against your baseline.
A good first use case is boring, frequent and easy to measure. Exciting ideas can come second.
Kamal Abraham, CEO
Decide what comes next
After the pilot, look at the numbers with the team. If the results are strong, roll the workflow out. If not, fix the weak points or move to the next idea on your list. Either way, you now have a repeatable process for every AI project that follows.