2 min readStrategy & first steps
How to choose your first AI project (without wasting a quarter)
Most first AI projects fail because they were the wrong project, not because the model was weak. A simple way to pick a workflow that is valuable, feasible, and measurable.
When a business decides to "do something with AI", the first idea is often the most visible one: a chatbot on the homepage, or an assistant that knows everything about the company. Those can work, but they're broad, hard to measure, and easy to stall.
A better first project is narrow, frequent, and already measured. Here's how I help teams find it.
1. List where time and money leak
Ask each team which tasks are repetitive, slow, or error-prone. Missed calls, unanswered enquiries, manual data entry between systems, chasing invoices, and answering the same internal questions all come up again and again.
2. Score each idea on three things
- Value: how often does it happen, and what does each occurrence cost in time, revenue, or customer experience?
- Feasibility: is the data available and accessible, and can the systems involved be connected?
- Risk: what happens when the AI gets it wrong, and can a person review before it matters?
The best first project scores well on value and feasibility, and has a low cost of mistakes. Drafting replies for a person to approve beats sending them automatically.
3. Measure the baseline before you build
Write down how the work is done today: how long it takes, how often it goes wrong, and what it costs. Without a baseline, you'll never be able to show the project worked, and the second project will be much harder to fund.
4. Pilot one workflow with real users
Build the smallest version that runs on real data with the people who do the work. Agree upfront what result would justify a rollout, and what would mean stopping. A clear "no" after four weeks is a good outcome too.
The goal of a first AI project is a confident decision, backed by evidence.
If you'd like a structured way to do this, the free AI readiness scorecard on this site walks through the same questions in about five minutes.