Scope to one persona, one workflow, one decision

A head of customer experience at a travel company told me about the pilot she is least proud of and learned the most from. The brief was to use AI to improve the customer experience. That was the whole scope. Improve the experience. Her team built something that tried to help agents, self-service customers, and the loyalty team all at once, and it helped none of them well enough to keep. It was too broad to test and too vague to measure. It quietly died.

The next one shipped. The difference was not the technology. It was that she cut the scope down to a single sentence with three blanks filled in. One persona. One workflow. One decision.

Narrow until it feels almost too small

Her rule now is to keep narrowing the scope until it feels slightly embarrassing, then stop. Not “help our agents.” One persona: a first-line agent handling a rebooking after a cancelled flight. Not “improve rebookings.” One workflow: the moment the agent has the passenger on the line and needs the next viable itinerary. One decision: which alternative flight to offer first.

When you get that specific, the whole project becomes testable. You know who the user is, so you can sit with three of them and watch. You know the workflow, so you know exactly where the AI shows up and where it does not. You know the decision, so you can compare the AI’s suggestion against what the agent would have picked anyway. Success stops being a mood and becomes a comparison.

Breadth feels safer in a planning meeting. It sounds ambitious, it flatters the budget, and it lets everyone see their own wish in the project. In delivery it is the opposite of safe. A broad pilot has no clear user to interview, no single flow to instrument, and no honest way to say whether it worked. You cannot measure “the experience.” You can measure whether the rebooking agent picked the AI’s first suggestion and whether the passenger accepted it.

The single decision is the anchor

She said the decision is the piece people skip, and it is the one that holds the rest together. A persona and a workflow still leave room to wander. Adding the single decision pins it down. Everything the feature does either helps make that one call better or it is out of scope for the pilot.

Here is the shape she uses now before any pilot starts:

One persona, named by role and the moment they are in, not a department.
One workflow, the specific sequence of steps that person is inside.
One decision, the single choice the AI is meant to change or speed up.
One comparison, what the person would have done without the AI.

If any of those four is plural, she sends the idea back. A friend who runs product at a logistics firm has the same instinct from the delivery side. He told me his teams ship AI pilots that fit on an index card and stall on the ones that need a slide deck to explain who they help. The index card ones can be measured in a week. The deck ones argue for a quarter.

Narrow scope is not a smaller ambition. It is how you find out whether the ambition is real before you spend on it. Once the one-persona pilot works and you can prove it, you widen to the second persona with evidence in hand instead of hope.

How we approach it at Density Labs

In the AI Opportunity Assessment, our two-week fixed engagement at $2,500, a lot of the work is subtraction. Teams arrive with a broad ambition, and we help them cut it to a single persona, a single workflow, and a single decision that can ship and be measured. That first slice is usually smaller than anyone expected walking in, and far more likely to survive contact with real users. Proving one narrow thing is what earns the right to build the wide version.

If your AI pilot helps everyone, you will not be able to tell whether it helped anyone. Pick one person, one flow, one decision, and make that work first.