How I AI“PMs who use AI will replace those who don’t”: Google’s AI product lead on the new PM toolkit
CHAPTERS
- 0:00 – 4:20
Becoming an AI‑enhanced PM: why the toolkit matters now
Marily Nika explains how her PM practice evolved from “pure AI PM” work with scientists to becoming an “AI‑enhanced PM” who uses modern AI tools daily. The framing sets up the episode’s core claim: AI changes not just what PMs build, but how they do product work end-to-end.
- •Shift from traditional PM workflows to AI-accelerated workflows
- •Goal: more impact, productivity, and faster iteration cycles
- •AI as augmentation: PMs who use AI outpace those who don’t
- •Episode promise: research → PRD → prototype → stakeholder-ready assets
- 4:20 – 6:29
Smart-fridge spark: a viral ‘expiring items’ moment as product insight
A surprising smart-fridge notification (including an “80-day-old Coca-Cola”) becomes the catalyst for a full product exploration. Marily uses the real-life moment and strong community reaction (likes/comments/reposts) to justify a rich opportunity space for busy families.
- •Unexpected fridge capability reveals latent product opportunity
- •Viral LinkedIn responses serve as early signal of interest
- •Identifying the gap: current experience (dismiss/snooze/recipes) is weak
- •Defining the lens: PM thinking applied to everyday devices
- 6:29 – 8:00
3-minute user research with Perplexity: mining Reddit at scale
Marily demonstrates fast market/user research by using Perplexity’s discussions filter to search Reddit opinions about smart fridges. She highlights how quickly you can surface sentiment, adoption barriers, and use cases—plus jump into source threads for context.
- •Use Perplexity ‘discussions’ filter to focus on Reddit opinions
- •Rapidly gather pros/cons, concerns, and adoption friction
- •Pull references and drill into original threads for nuance
- •Treat this as a high-leverage starting point, not final research
- 8:00 – 11:17
Forcing critical thinking: pro/anti agents debate to find the minimum lovable feature set
To avoid overly agreeable AI outputs, Marily prompts Perplexity to create two agents—one pro and one against smart fridges—and have them debate multiple rounds. The goal is a concrete, PM-usable output: the minimum feature set needed to convince skeptics and move toward product-market fit.
- •Combat “agreeableness” by generating opposing personas
- •Multi-round debate surfaces real objections and stronger counterarguments
- •End goal is feature requirements, not just analysis
- •Output becomes structured input for downstream tooling
- 11:17 – 14:06
Tool hopping on purpose: converting research into a PRD in ChatGPT
Marily copies the minimum feature set from Perplexity into a custom “AI Product GPT” that generates a PRD in her preferred structure and voice. The PRD provides a strong head start—reducing busywork and enabling more time for strategic thinking and iteration.
- •Custom GPT with PRD template + personal writing style
- •Paste features → generate PRD with problem, users, prioritization, architecture
- •PRD is a draft accelerator (not ‘perfect on first try’)
- •Structure + content constraints improve output quality
- 14:06 – 17:00
From PRD to interactive UI: prototyping in v0 with better inputs
With the PRD as a high-quality spec, Marily uses v0 to generate an interactive smart-fridge dashboard UI. She emphasizes experimentation and concludes that investing in better upfront inputs reduces exhausting downstream tweaking.
- •Paste full PRD into v0 to generate UI quickly
- •PRD-driven prompting yields higher fidelity prototypes
- •Iterative mindset: fewer later tweaks when starting with strong inputs
- •Prototyping tools mentioned: v0, Lovable, Google AI Studio
- 17:00 – 21:31
Prototypes as influence: winning product reviews with something stakeholders can click
Marily argues prototypes are most powerful in product reviews, where faster AI cycles mean more frequent scrutiny and more skepticism. A clickable prototype communicates vision, credibility, and value more effectively than a PRD-turned-deck, especially for resource/investment decisions.
- •Best use case: product reviews and stakeholder alignment meetings
- •Clickable experience beats slides for communicating vision
- •Prototype increases credibility and speeds decision-making
- •AI-era cadence: more reviews, faster cycles, higher need for compelling artifacts
- 21:31 – 30:16
Adding video to sell the vision: Flow/Veo and Sora for promo clips
To help people experience the product from a user perspective, Marily generates promotional videos from text using Google Labs’ Flow (Veo) and compares it with OpenAI’s Sora. The segment shows both the power and the quirks (e.g., weird screens, continuity issues) of current video generation.
- •Text-to-video as a stakeholder persuasion layer beyond prototypes
- •Flow setup: prompt + model options (quality, aspect ratio, audio)
- •Sora comparison: faster generation, cameos, social/distribution layer
- •Reality check: artifacts (odd UI placements, character inconsistencies) still happen
- 30:16 – 32:22
The 20-minute end-to-end PM workflow: research → PRD → prototype → video
Claire zooms out to recap how the team produced a full product package in minutes: Reddit-based research, defensible features, a PRD, a clickable prototype, and a video asset. The conclusion: AI flips product management on its head—changing how PMs build, communicate, and influence.
- •End-to-end workflow can be ~15–20 minutes without commentary
- •Outputs become a compelling ‘package’ for internal persuasion
- •AI changes communication and influence as much as execution speed
- •Core thesis: AI-equipped PMs replace those who don’t adopt
- 32:22 – 37:23
NotebookLM as an AI judge: scoring demo day pitches via audio overviews
Marily shares a real-world operational use case: using NotebookLM to judge her bootcamp’s demo day by uploading pitch audio and generating rankings based on innovation, impact, and storytelling. The AI results align with human judges, and the interactive “radio call-in” mode adds drama and fun.
- •Upload multiple pitch recordings into a NotebookLM notebook
- •Generate audio overview with explicit judging criteria
- •Interactive mode enables ‘call-in’ style Q&A (e.g., ‘just tell me the winners’)
- •Use cases extend to hackathons, pitch competitions, sales enablement events
- 37:23 – 38:36
When tools fail: reset, use AI to write better prompts, and get specific
In the lightning round, Marily explains her approach when generation quality is off: kill the attempt and start over. She recommends using AI to improve prompts and making them longer and more detailed to reduce iteration cycles.
- •Don’t fight a bad generation—restart the instance
- •Use AI to help craft the prompt (‘AI on AI’)
- •Longer, more detailed prompts reduce back-and-forth iterations
- •Practical mindset: speed comes from fewer cycles, not rushed inputs
- 38:36 – 40:09
Wrap-up: where to find Marily and her AI Product Academy bootcamp
The episode closes with where to follow Marily (LinkedIn) and how to engage with her bootcamp and certification program. Claire wraps with show distribution details and a call to like/subscribe/review.
- •Find Marily on LinkedIn for AI product management + building content
- •AI Product Academy bootcamp: teams build end-to-end with assigned engineers
- •Mentions upcoming cohort timing and certifications
- •Show outro: subscribe, comment, listen on podcast platforms
