Strategy
Scoping an AI MVP Around One Complete User Journey
Plan an AI MVP with a clear user journey, useful boundaries, and reviewable results so the first release can answer a concrete product question.
A minimum viable product is still a product: a person should be able to start a task, receive a useful result, and understand what to do next. For AI features, an impressive output alone does not establish that journey. Scope the first release around a complete, modest workflow rather than a collection of disconnected demonstrations.
Write the journey in ordinary language
Describe who arrives, what information they bring, and what result lets them move forward. A hypothetical property manager might upload a maintenance request, receive a suggested category, and approve its assignment. That journey includes intake, a suggestion, review, and a saved decision.
Decide which steps need custom development and which can remain manual during the trial. Manual operations are reasonable when they are visible to the team and sustainable at the planned scale. Avoid silently depending on someone to repair outputs without recording that effort; it distorts what the MVP teaches you.
Separate essential behavior from later convenience
List every proposed feature beside the user problem it addresses. Keep the capabilities required to complete the chosen journey. Move additional channels, advanced personalization, multiple user roles, and elaborate reporting into a later list unless they are genuinely required by that first audience.
Treat failure handling as part of the essential journey. Users need a way to correct input, see that a task is still running, retry an interrupted action, and recover their work. A smaller feature set with these basics can be more informative than a broad prototype that works only under ideal conditions.
Define the decision the MVP will inform
Choose an explicit evaluation question, such as whether staff accept suggested categories after checking the request. Observe rejected suggestions and time spent reviewing, alongside completed tasks. A usage count alone cannot show whether the feature makes the workflow easier.
Set a review date and name the people who will interpret the evidence. Decide in advance what would justify extending the feature, changing its boundaries, or stopping it. The MVP should produce a product decision as well as a working interface, and its scope should support that decision.
