(Product experience)
/Improving AI assistant outcomes
Used conversation evidence, product analytics, and customer feedback to improve assistant behavior and prioritize product work.
- status
- Product experience
- scope
- Discovery, analytics, PRDs, prioritization, experimentation, and GTM collaboration
- year
- 2025 – Present
1,500+
conversations analyzed
50+
ad variants supported
AI
assistant and growth workflows
(01)
/the opportunity.
AI assistant quality appears in the conversations that succeed, stall, or fall back. The team needed a clearer way to turn those signals into decisions across assistant behavior, onboarding, and proactive experiences.
(02)
/my approach.
- Reviewed 1,500+ conversations and grouped failure patterns, unmet intent, and points where context broke down.
- Combined qualitative findings with Mixpanel analysis and support feedback to shape priorities, PRDs, and sprint planning.
- Worked with product and engineering partners on context-aware behavior, fallback reduction, onboarding, and proactive experiences.
- Supported AI workflows that generated 50+ ad creative variants plus paid search and content experiments.
(03)
/what changed.
The work contributed to better successful-user outcomes and lower fallback behavior. No exact percentage is stated because the available evidence does not provide a reliable baseline or attribution model.
(04)
/what I learned.
Conversation analysis becomes useful when it changes a backlog, an experiment, or a product rule. Connecting a pattern to an owner and a decision is what turns research activity into product work.
(resources)