Aakash GuptaI got a private Masterclass in AI PM from Google AI PM Director
CHAPTERS
- 0:00 – 1:55
Why AI PM is a real (and fast-growing) role at Google
The episode opens by framing Google’s momentum in AI and sets up the promise: a practical masterclass in AI product management plus hands-on demos of Google’s AI tools. Jacqueline confirms the AI PM role is real and anchors the conversation in building AI-native products.
- •Google’s perceived shift from behind to leading in AI models and tooling
- •What an AI PM does: shipping AI-based and AI-native products
- •The episode’s structure: demos + frameworks + hiring/interview guidance
- 1:55 – 2:59
AI PM levels & compensation: calibrating Google vs startup expectations
Aakash and Jacqueline discuss how pay and leveling work for Google PM roles, and why “senior PM” can mean very different things across companies. Jacqueline shares her own leveling history to illustrate how experience and scope map to Google’s ladder.
- •Compensation is generally strong; compare via official postings and market data
- •Google PM leveling differs from startup titles; context matters
- •Typical entry levels and how progression works with experience/ships
- 2:59 – 4:16
How AI changes product building: faster tools, new product possibilities, and overwhelm
Jacqueline explains two major shifts: AI changes how products are built (tools, speed) and what products are worth building (capabilities unlocked by models). She highlights the emotional reality—excitement and overwhelm—due to rapid model/tool evolution.
- •AI transforms both the build process and the feature/product space
- •Pace of change creates both opportunity and cognitive overload
- •More powerful models → better tools → faster shipping and more competition
- 4:16 – 6:10
Zero-to-one in the AI era: embracing discomfort and bringing clarity to chaos
Jacqueline reframes messy early-stage building as both fun and confusing, emphasizing that discomfort is often a signal of meaningful innovation. She shares a mindset shift: don’t confuse “uncomfortable” with “wrong,” especially when doing true 10x thinking.
- •Zero-to-one is inherently messy; AI adds sparkle but also ambiguity
- •Discomfort can indicate you’re pushing into real innovation territory
- •Strong PMs create focus, align teams, and move forward through uncertainty
- 6:10 – 9:06
Nano Banana capabilities tour + Demo: colorizing and restoring old photos
Jacqueline showcases Nano Banana’s image-editing range (rotation, annotations, sketch-to-art, seasonal inference) and then demonstrates photo restoration by colorizing her grandparents’ wedding photo. The segment stresses iteration and careful prompting to get high-quality outputs.
- •Nano Banana can edit, transform, and infer context (e.g., realistic winter scenes)
- •Old-photo restoration/colorization workflow and why it’s compelling
- •Prompt iteration is essential; keep refining when results miss the mark
- 9:06 – 10:14
Prompt engineering workflow: using Gemini to refine prompts + negative prompts
Aakash asks for a prompt breakdown, and Jacqueline explains how she uses Gemini Pro to improve prompts when outputs aren’t right. She explains prompt structure (color goals, lighting, texture, lens optics) and the role of negative prompts to prevent common failure modes.
- •Meta-prompting: ask Gemini to diagnose what went wrong and suggest edits
- •Prompt components: desired aesthetic, lighting, realism/texture, camera optics
- •Negative prompts help remove recurring artifacts and undesired styles
- 10:14 – 12:54
Access paths & Mixboard: a canvas UI for multimodal brainstorming with Nano Banana
Jacqueline explains where to use Nano Banana (AI Studio, Gemini app) and introduces Mixboard as a canvas for ideation and batch transformations. She demos style transfer and multi-image operations (e.g., converting several images into sketches).
- •Ways to access Nano Banana: AI Studio, Gemini app, and Mixboard
- •Mixboard differs from chat: open canvas for visual storytelling and ideation
- •Batch editing and style transfer workflows powered by Nano Banana
- 12:54 – 16:09
Demo: pet photo → drone show image → Veo video (and how to keep scenes consistent)
Jacqueline demos turning a pet photo into a drone-show-style image, compares Gemini vs AI Studio output quality, then chains the result into Veo to generate a video. She shares tips on consistency using Nano Banana as a “seed” image and hints at tools like Flow for multi-scene construction.
- •Quality can vary by surface (Gemini app vs AI Studio); try both
- •Chaining: image generation/editing → video generation in Veo
- •Consistency strategy: iteratively generate a consistent character/scene via Nano Banana, then use as Veo seed
- 16:09 – 20:16
Ads break: Vanta and Pendo sponsorships
A sponsored segment covering security/compliance automation (Vanta) and product analytics plus AI agents for SaaS experiences (Pendo).
- •Vanta positioning: faster audits + continuous monitoring for fast-moving teams
- •Pendo positioning: AI agents that act on product data with measurable ROI
- •Sponsor URLs and calls to action
- 20:16 – 27:35
Opal demo: building mini AI apps via prompt chains (workflows, sharing, remixing)
Jacqueline introduces Opal as a tool to build, edit, and share mini AI apps using natural language, then walks through examples including an image-collage app and a storybook maker. She demonstrates generating a resume-critique app from a URL, highlights model swapping, and emphasizes share/remix as distribution mechanics.
- •Opal = chained prompts + user inputs + configurable outputs (Docs/Sheets etc.)
- •Examples: nature collage workflow; storybook generator; resume critique from a blog URL
- •Shareable/remixable apps accelerate iteration and adoption; model choice matters
- 27:35 – 29:45
Building in public & side-project strategy: stress-testing ideas before going “production”
Jacqueline explains when to prototype in tools like Opal/AI Studio vs building a full production app or agent. She advocates building in public, collecting feedback early, and maintaining multiple side projects to explore possibilities and develop better product instincts.
- •Prototype first to validate the idea; don’t over-invest before conviction
- •Side projects broaden creativity, surface new use cases, and build a portfolio
- •Early feedback loops (public sharing) help refine direction and taste
- 29:45 – 39:39
Agent product frameworks: anatomy, interaction spectrum, and the inverted triangle MVP
Jacqueline shares three core frameworks for AI/agent product thinking: (1) anatomy of an agent (models, tools, memory), (2) user interaction spectrum (do-it-for-me vs do-it-with-me), and (3) inverted triangle (think big, ship small via scope/positioning/audience). She emphasizes tool use (APIs/UI actions, MCP) and memory/personalization as key design pillars.
- •Anatomy of an agent: model capabilities + tool use + memory/personalization
- •UX spectrum: autonomous agents vs collaborative copilots changes design needs
- •Inverted triangle: big vision, MVP scope cuts, beta positioning, staged rollout via audience
- 39:39 – 48:13
Choosing the right problems: paradigm shifts, future-proofing, and first-principles thinking
Jacqueline offers decision questions to ensure you’re building something durable: are you creating a new workflow (car) or incremental improvement (faster horse)? She explains how model advances can commoditize features and why teams must be willing to drop obsolete work, grounding the approach in first-principles and second-order/platform thinking.
- •Paradigm shift question: new workflow vs incremental optimization
- •Future-proofing: anticipate model upgrades that commoditize or unlock capabilities
- •First principles: focus on the true user need, not legacy tool patterns; be willing to discard outdated work
- 48:13 – 1:04:26
Hiring an AI PM at Google: the 6 characteristics + resume, interview, and 18-month roadmap
Jacqueline outlines what she seeks in AI PM candidates—product taste, systems thinking, clarity in chaos, storytelling, ownership, and AI intuition/creativity—then gives actionable resume guidance (brevity, links, context, side projects). She discusses interview expectations (product thinking over pure vibe coding), describes her hiring loop variant, and closes with an 18-month plan centered on building, networking, learning, and sharing work publicly.
- •Six hiring signals: taste, vision/systems, chaos-to-clarity, storytelling, ownership, AI intuition/creativity
- •Resume advice: one page, specific evidence/links, thoughtful design, contextualized impact, meticulous proofreading
- •Interview guidance + roadmap: lead with product framing, build artifacts, network, immerse in the space, and publish learnings
