How to choose your first AI use case
· 5 min read · Hidden Node
Most first AI projects fail for a boring reason: the wrong problem was chosen. The model worked, but the workflow did not change, nobody measured anything, or the data was locked away. Picking well matters more than picking fast.
Start from work, not from technology
List the tasks your teams repeat every day. Good candidates usually involve reading something (emails, PDFs, forms, tickets), making a judgement against rules, and then updating a system or replying to someone. Claims triage, invoice matching, KYC review and support tickets all fit this shape.
Score each candidate on five questions
- Volume. Does it happen hundreds or thousands of times a month? Small volumes rarely justify the build.
- Rules plus judgement. Is there a clear standard for a good outcome, even if applying it takes skill? If experts disagree on every case, AI will struggle too.
- Data access. Can we actually get the inputs (documents, records, history) without a six-month integration project?
- A number that moves. Can you name the metric that should change: hours per case, backlog size, turnaround time, error rate?
- Safe to get wrong. If the system makes a mistake, is it caught before it hurts a customer? Early projects should keep a person approving the output.
Give each a score from 1 to 3. The highest total is usually your first project.
Watch for these traps
- The demo trap. A chatbot that answers general questions looks impressive and changes nothing. Tie the project to a real queue of work.
- The everything trap. "Automate customer service" is a programme, not a project. Pick one slice, such as order-status questions.
- The clean-data trap. Waiting for perfect data can take years. Many good projects start with documents and email exactly as they are today.
What good looks like
A strong first use case reaches real users within a few months, has a person reviewing outputs at the start, and has a baseline measured before launch so the improvement is visible. Once it works, the same foundations (data access, evaluation, monitoring) make the second project much faster.
If you would like a second opinion on your shortlist, our free readiness assessment does exactly this exercise with your team.
Want help applying this in your organization?
Get a free readiness assessment