Pick the flow that matters most — onboarding, checkout, your AI copilot, the triage funnel — and I take it apart the way a real user experiences it: where it hesitates, over-explains, dead-ends, or quietly loses people.

The AI-native lens comes standard. If your product has chat, voice, or agent surfaces, that's my home turf: turn structure, disclosure moments, hallucination surfaces, errors that read as failures instead of forks, human handoff. Most teams shipping LLM features get the model right and the interaction wrong — that gap is where I work.

Walk your flow as a real user would — every path, including the ones you'd rather I didn't take

Map the interaction architecture — what the system asks, when it asks it, what it does with a vague or hostile answer

Stress the failure modes — empty states, dead ends, silent errors, cold starts, the moment a user loses the thread

Score against principles fit for the system — probabilistic-UX principles for AI surfaces, not just generic usability heuristics

Prioritise — what to fix this sprint, what to fix next quarter, what to stop doing entirely

What you get back: a findings document you can act on immediately, a short recorded walkthrough so your team hears the reasoning rather than just reading verdicts, and a call to argue with me about it.

What this is not: a visual design pass or an implementation engagement. It's a focused diagnostic on one flow, delivered fast — auditing the whole journey is the UX Audit — Full Product ($850).

I design AI-native products end to end — conversational triage, agent interfaces, LLM-assisted flows — and I build them, not just draw them. That means the fixes I recommend are ones that can actually ship.