How I design AI products.

Designing for AI is fundamentally different from designing for deterministic software. Here are the core patterns I work with.

1. LLM output patterns

At Geofy, I designed an interface where AI-generated prospect dialogue led users into a screening interface. The core challenge was that AI output varied — sometimes short, sometimes long, sometimes off. I had to design for robustness against variable output: truncation, overflow, empty states, and unexpected formatting. That's the fundamental challenge with any LLM-driven UI — you can never assume what the output looks like.

2. Feedback loops

At OwnU, I designed thumbs up/down mechanics on AI-generated workout suggestions that fed directly into the next iteration of suggestions. For Lyyti, the equivalent would be an event organiser being able to correct or adjust AI-generated communication drafts inline — before they go out to thousands of participants. The feedback mechanism has to feel natural, not like a form.

3. Latency as a UX constraint

LLM calls take 2–10 seconds. That fundamentally changes interaction design. You can't have a button that just "loads." For Lyyti this could mean designing an AI feature that analyses participant data — instead of a spinner, you show "Analysing registration patterns…" with progressive steps. The wait becomes part of the experience rather than dead time.

4. Human-in-the-loop mechanics

AI should never send a communication or make a decision on behalf of an event organiser without a clear review step. Especially in Lyyti's context — one wrong AI-generated email going out to 5,000 participants is a serious problem. The design challenge is making the review step feel lightweight and natural, not like an extra hurdle. Clear "Edit", "Approve" and "Override" actions, always visible confidence signals when the AI is uncertain.

5. A combined example: Geofy / Elvy

At Geofy I designed a full AI-driven lead qualification flow where AI handled the initial prospect dialogue, steered leads through qualification steps, and handed them off to a human screening interface at the right moment. That handoff moment — designing what the human sees, what the AI has collected, and where human judgment takes over — is exactly the core challenge in any B2B SaaS AI feature. It combines all four patterns: variable output, feedback, latency and human-in-the-loop in one flow.