Aakash GuptaHow To ACE AI Product Sense Interviews (OpenAI PM Mock Interview)
CHAPTERS
Why AI Product Sense interviews are suddenly table-stakes
Aakash frames a gap in existing interview prep: plenty of generic product-sense content, but almost nothing tailored to AI Product Sense—now a common case style at top AI labs. He previews the format (a 45-minute “speed run” of PM thinking) and what viewers will get from the video.
The case prompt: double WAU for ChatGPT image creation with 3 engineers
Dr. Bart reveals the interview question: grow weekly active users of ChatGPT’s image creation from 175M to 350M in just three months with only three engineers. They quickly align on what counts as “image creation” (still images in ChatGPT, not Sora video).
Clarifying product boundaries, teams, and what “3 engineers” really means
Aakash clarifies how the image feature fits within OpenAI’s broader multimodal strategy and distinguishes model research from product engineering. He sets expectations on what’s feasible: influence research priorities but primarily ship product/UX changes with a small engineering team.
A live framework: mission → users → problems → solutions → metrics → guardrails
Aakash builds a six-part framework on a Miro board to structure the case and keep the interviewer aligned. The framework is designed to be collaborative and adaptable as new information emerges during the interview.
User segmentation and growth math: where could +175M WAU come from?
He proposes a rough segmentation of current image-gen users and uses ChatGPT’s overall WAU to reason about penetration. The conclusion: most incremental growth must come from lower-tech-literacy users, not early adopters.
Diagnosing user problems by bucket: UI flow, app functionality, and model constraints
Aakash organizes pain points into three buckets and does a quick “use the product” UX walkthrough to surface friction. He highlights discoverability, confusing/long loading behavior, and missing creation/editing capabilities that block broader adoption.
Competitor scan: what Midjourney, Runway/NanoBanana, and Synthesia teach
Prompted by the interviewer, he compares key competitors to extract actionable gaps. The discussion distinguishes where competitors win (quality/styles, editing workflows, avatars/likeness) and maps those insights back into the problem list.
Understanding user jobs-to-be-done via prompt and feedback data
Aakash explains how he’d empirically determine what users are trying to accomplish: cluster prompt intent, measure satisfaction, and identify weak areas to improve. He ties this to prioritization signals such as thumbs up/down by use-case category.
Solution brainstorming: improve discoverability, loading delight, and editing workflows
He rapidly ideates solutions across UI and functionality, borrowing from gaming and productivity UX patterns. The emphasis is on quick wins that help low-tech users succeed and return, not only on model improvements.
Prioritization under constraints: impact vs effort to hit 350M in 3 months
Aakash shifts to sizing and sequencing initiatives given only three engineers. He argues UI/discoverability changes are fast and could deliver large adoption gains, while larger bets (standalone app, deep personalization) are deprioritized due to time/effort.
Curveball: build an Instagram killer in 3 months—MVP strategy
Dr. Bart forces an alternate scenario: ship a new image-focused app quickly. Aakash proposes an MVP centered on AI photo editing for authenticity, filters, and practical enhancements (zoom out, background fill, resolution), targeting a narrow user segment for viral growth.
Final plan details: editors for memes/thumbnails/infographics + success metrics logic
He consolidates a shortlist: UI improvements, editing features, and purpose-built canvases for common outcomes like thumbnails, profiles, memes, and infographics. He sanity-checks rough user lift vs timeline, aiming to bridge the additional 175M WAU.
Safety, copyright, and policy: guardrails for editing and viral styles
Aakash closes by explicitly addressing risks: likeness abuse, harmful content, and copyright conflicts (e.g., Ghibli-style controversy). He recommends aligning with leadership but leans toward being copyright-safe early to avoid backlash and low reviews.
Debrief: interviewer feedback + three interview takeaways to replicate
Dr. Bart praises the end-to-end discovery journey, adaptability to curveballs, and clear narrative via the Miro board, with minor notes on being more explicit about validating estimates and internal research. Aakash then distills three tactical takeaways for candidates: build a unique framework, check in and adapt live, and sell your unique fit throughout the case.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome