Your AI feature shipped. Adoption is soft, the support tickets are strange, and nobody on the team can quite explain why users do not lean on it.

That is usually not a model problem. It is an interface problem. The product never made the AI legible, steerable, or safe to be wrong in front of.

A structured two week review of your AI surfaces against established usability heuristics plus the criteria that only matter for AI products: is the model's behavior legible, is the user in control, are failures honest, and is there a way back when it gets something wrong.

You get findings tied to your actual screens. Each one is prioritized and paired with a specific fix.

Week 1. Kickoff call to understand your users, your model, and where you think the pain is. I then work through the product the way your users do, documenting friction against the heuristic set.

Week 2. I score and prioritize every finding, annotate the screens, and write the report. We close with a working session where I walk your team through it and we agree what gets fixed first.

Not "improve clarity." Each finding names the screen, states the problem, cites the heuristic or AI-specific criterion it breaks, rates it P0, P1, or P2, and recommends a specific change. Written so a PM or an engineer can put it straight into a ticket without a designer translating first.

Teams who have shipped an AI feature and are seeing weak adoption, low trust, or support load they cannot explain. It works whether you have a design team or none at all.

Concepts with nothing built yet. If you are still deciding what the experience should be, the Conversational and Agentic UX Design Sprint is the better starting point.