Aakash GuptaI got a private Masterclass in AI PM from Google AI PM Director
Aakash Gupta and Jaclyn Konzelmann on google AI PM Director demos tools and shares AI PM playbook.
In this episode of Aakash Gupta, featuring Jaclyn Konzelmann and Aakash Gupta, I got a private Masterclass in AI PM from Google AI PM Director explores google AI PM Director demos tools and shares AI PM playbook AI has changed product building both by accelerating how teams prototype/ship and by expanding what products can do, while also making the landscape feel fast-moving and overwhelming.
At a glance
WHAT IT’S REALLY ABOUT
Google AI PM Director demos tools and shares AI PM playbook
- AI has changed product building both by accelerating how teams prototype/ship and by expanding what products can do, while also making the landscape feel fast-moving and overwhelming.
- Konzelmann demos practical, high-leverage workflows with Google tools—Nano Banana for image editing, Veo for turning images into videos, and Mixboard/Opal for multimodal creation and prompt-chaining mini-apps.
- She argues AI PM is a real role and emphasizes “thinking big” and first-principles product work so teams don’t build features that models will soon commoditize.
- Three frameworks guide AI product decisions: an agent’s anatomy (models, tools, memory), a user-interaction spectrum (do-it-for-me vs do-it-with-me), and an inverted triangle to ship MVPs while preserving a large vision.
- Hiring and career advice centers on product taste, systems thinking, clarity in chaos, storytelling, execution ownership, and AI intuition—backed by tangible public side projects and crisp, link-rich resumes.
IDEAS WORTH REMEMBERING
8 ideasTreat discomfort as a signal of real 10x thinking, not a sign you’re wrong.
Konzelmann notes zero-to-one AI work is inherently messy and uncomfortable; separating “uncomfortable” from “wrong” helps teams push through ambiguity and innovate rather than retreat to safe iterations.
Prompting is an iterative product skill—use models to improve your prompts.
She repeatedly uses Gemini to critique and refine prompts (including negative prompts) when outputs miss the mark, turning prompt iteration into a systematic loop rather than guesswork.
Use the right surface for the job: chat, canvas, workflow, or agent.
Gemini/AI Studio are great for quick exploration, Mixboard supports open-ended visual ideation, Opal packages repeatable prompt chains into shareable mini-apps, and full agents add tool use and memory for autonomous task completion.
For scene consistency, generate a consistent ‘seed’ image first, then animate it.
She keeps a character consistent across images with Nano Banana, then uses those images as starting frames for Veo video generation—reducing drift across short vignettes.
Design agent experiences along the “do it for me” ↔ “do it with me” spectrum.
Deep research and batch generation fit autonomous modes with minimal interruption, while vibe coding and interactive audio require collaborative loops; picking the wrong mode breaks user expectations.
Future-proofing is mandatory: assume models will commoditize today’s differentiators.
She urges PMs to ask how the next model update affects strategy—whether it erases your core feature or unlocks new capabilities—so you build on tailwinds rather than get ‘crushed by the wave.’
To ship big ideas fast, narrow scope, label honestly, and limit the audience.
Her inverted-triangle approach ships MVPs by cutting features, positioning launches as betas/experiments, and starting with trusted testers—while keeping the long-term vision intact.
What Google looks for in AI PMs is broader than “AI knowledge.”
Her six characteristics prioritize product taste, systems thinking, leading through ambiguity, storytelling without perfect data, full-spectrum ownership, and applied AI creativity—often evidenced through public side projects.
WORDS WORTH SAVING
5 quotesIt does really feel like there's never been a more exciting time to build than right now.
— Jaclyn Konzelmann
True 10X thinking is actually supposed to feel uncomfortable… I kept confusing uncomfortable with wrong.
— Jaclyn Konzelmann
Learning how to prompt these models is a skill that takes time but is incredibly worth it.
— Jaclyn Konzelmann
Are you just building a faster horse or are you building something net new, like a car?
— Jaclyn Konzelmann
If the first thing you do is jump straight into execution mode, that would worry me.
— Jaclyn Konzelmann
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsNano Banana vs Gemini vs Mixboard: what concrete tasks perform best on each, and why do outputs sometimes differ across surfaces?
AI has changed product building both by accelerating how teams prototype/ship and by expanding what products can do, while also making the landscape feel fast-moving and overwhelming.
In your photo-colorization demo, which parts of the prompt mattered most (lighting, optics, saturation, negative prompts), and what’s your repeatable method for prompt debugging?
Konzelmann demos practical, high-leverage workflows with Google tools—Nano Banana for image editing, Veo for turning images into videos, and Mixboard/Opal for multimodal creation and prompt-chaining mini-apps.
For Veo image-to-video, what are the top tactics to extend beyond 8-second vignettes while maintaining character and environment consistency?
She argues AI PM is a real role and emphasizes “thinking big” and first-principles product work so teams don’t build features that models will soon commoditize.
Opal seems positioned between Zapier/n8n and agent builders—what use cases do you believe will remain ‘workflow apps’ versus evolving into autonomous agents?
Three frameworks guide AI product decisions: an agent’s anatomy (models, tools, memory), a user-interaction spectrum (do-it-for-me vs do-it-with-me), and an inverted triangle to ship MVPs while preserving a large vision.
You mentioned MCP, APIs, and UI actions (e.g., Project Mariner). How should PMs choose between API tool use vs UI automation for reliability and safety?
Hiring and career advice centers on product taste, systems thinking, clarity in chaos, storytelling, execution ownership, and AI intuition—backed by tangible public side projects and crisp, link-rich resumes.
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
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