Empathize: Human First, AI Second

I start every project by stepping into users’ shoes - conducting interviews, shadowing workflows, and mapping emotional journey points. In the AI era, I pay special attention to users’ hopes and fears around automation: What tasks do they want to offload? Where do they worry AI might overstep? These insights shape the scope of AI features and guardrails before a single line of code is written.

Define: Framing AI Opportunities

With empathy data in hand, I craft clear problem statements that bridge business goals and user needs — e.g., “How might we use AI to cut research time in half without sacrificing trust?” By anchoring each “How might we” question in real pain points and success metrics (OKRs), we ensure AI isn’t a gimmick but a strategic lever.

Ideate: Co-Creating with AI in Mind

During brainstorming workshops, I invite stakeholders to sketch both human and AI “actors” in the flow. We explore where AI can suggest, automate, or augment, then surface potential risks—bias, hallucination, loss of control. This dual focus helps our ideas stay bold yet grounded in user control and transparency.

Prototype: Simulating Smart Interactions

I leverage Wizard of Oz prototypes and low-fi mockups that simulate AI responses. This allows rapid testing of suggestion cards, confidence cues, and override controls. By validating interaction patterns early, we avoid costly iterations on model training before we know what flows truly resonate.

Test: Measuring Trust and Usability

In usability sessions, I observe how users interpret AI suggestions, how often they accept or edit them, and how confident they feel. I pair qualitative feedback (“I’m not sure why it suggested this”) with quantitative metrics (acceptance rate, time-to-decide) to pinpoint UX friction and trust gaps.

Iterate: Continuous Learning Loop

Post-launch, I embed telemetry and in-product feedback (thumbs up/down, corrections) to capture real-world behavior. Weekly tuning sessions bring together UX, data science, and product teams to refine prompts, retrain models, and adjust UI affordances - keeping our design thinking cycle alive as the AI learns.