
Product Talk The 4-Step AI Prompt Framework Every Product Manager Should Know
10 snips
Apr 3, 2026 Akriti Dokania, venture partner and former engineer and product manager at Amazon and Microsoft, shares a PM-minded take on prompting AI. She discusses leading with context, using clarifying questions, assigning roles to the AI, naming risks, and starting with one small automation. Conversations cover what makes builders AI-native and the production challenges of reliability and privacy.
AI Snips
Chapters
Transcript
Episode notes
PM Mindset Is Output First
- Product managers think output-first and customer-impact-first rather than input or implementation details.
- Akriti contrasts PMs' truth-seeking customer focus to engineers' input/logical focus, which fits how LLMs respond to outcome-driven prompts.
Start Prompts With Rich Context
- Lead prompts with rich context (user state, constraints, preferences) instead of terse commands to get tailored, useful outputs.
- Akriti's wedding-assistant example shows adding bride status, cultural preferences, guest count, and vibe yielded relevant vendor suggestions and follow-up questions.
Wedding Assistant Agent Example
- Akriti built a wedding assistant agent that researched vendors, handled outreach, and narrowed choices for couples.
- She fed context like being a busy bride, South Asian preferences, multiple states, and got specific vendor research plus follow-up questions.
