“PMs who use AI will replace those who don’t”: Google’s AI product lead on the new PM toolkit

“PMs who use AI will replace those who don’t”: Google’s AI product lead on the new PM toolkit

How I AIDec 1, 202540m

Marily Nika (guest), Claire Vo (host)

AI-enhanced PM vs. traditional PM workflowReddit mining for fast user research (Perplexity)Persona/agent debate to reduce LLM agreeablenessPRD generation with a custom ChatGPT templatePRD-to-prototype via v0 (and similar tools)Prototypes and video as stakeholder influenceNotebookLM as an AI judge for demo daysPrompt recovery when tools fail (reset + prompt with AI)

In this episode of How I AI, featuring Marily Nika and Claire Vo, “PMs who use AI will replace those who don’t”: Google’s AI product lead on the new PM toolkit explores an end-to-end AI workflow that upgrades modern product management practice Marily Nika walks through how she “tool-hops” across AI products to compress a traditional product cycle into ~15–20 minutes: mine user sentiment, synthesize requirements, generate a PRD, build a clickable prototype, and produce a promo video for stakeholder influence.

An end-to-end AI workflow that upgrades modern product management practice

Marily Nika walks through how she “tool-hops” across AI products to compress a traditional product cycle into ~15–20 minutes: mine user sentiment, synthesize requirements, generate a PRD, build a clickable prototype, and produce a promo video for stakeholder influence.

She uses Perplexity’s Reddit/discussions search plus a pro/contra agent debate to surface adoption barriers and the minimum feature set needed to reach product-market fit for a smart-fridge concept.

She then turns those features into a structured PRD via a custom ChatGPT PRD generator and feeds the PRD into v0 to create an interactive UI prototype tailored to the requirements (e.g., local-first privacy).

Finally, she shows how video generation (Google Flow/Veo and OpenAI Sora) can communicate the product narrative, and how NotebookLM can act as a surprisingly effective multi-file “AI judge” for selecting winners at a demo day.

Key Takeaways

Use opinion-rich sources to approximate rapid user research.

By searching Reddit discussions in Perplexity, Marily quickly surfaces real-world use cases, concerns, and adoption barriers—then drills into cited threads when needed for context.

Get the full analysis with uListen AI

Force critical thinking with pro/contra agents to counter LLM agreeability.

Having two agents debate (one for, one against) creates a structured adversarial review and ends with a “minimum convincing feature set” aimed at winning over skeptics.

Get the full analysis with uListen AI

Generate outputs in the format required by the next step.

Marily explicitly asks for “features” because that becomes the input to PRD generation and prototyping; chaining tools works best when each step produces the next step’s ideal artifact.

Get the full analysis with uListen AI

A strong PRD is leverage, not paperwork, when feeding builder tools.

She finds that spending extra time improving the PRD reduces downstream prototyping tweaks, making the overall cycle faster and less exhausting than “prompting directly.”

Get the full analysis with uListen AI

Clickable prototypes outperform decks for product reviews and alignment.

In fast-moving AI orgs where skepticism can be high, a hands-on prototype increases credibility and makes the vision tangible, improving stakeholder buy-in during product reviews.

Get the full analysis with uListen AI

Video generation can sell a narrative—but will drift without clear prompts.

Flow/Veo outputs showed odd failures (e. ...

Get the full analysis with uListen AI

AI judging is viable when you provide clear criteria and consistent inputs.

NotebookLM can ingest many pitch recordings and produce a ranked selection using criteria like innovation, impact, and storytelling—useful for bootcamp demo days, hackathons, or pitch contests.

Get the full analysis with uListen AI

When generation goes off the rails, restart and let AI write the prompt.

Marily’s “kill the instance and start over” approach, combined with asking AI to draft a longer, more detailed prompt, reduces iteration loops and improves output fidelity.

Get the full analysis with uListen AI

Notable Quotes

“Within minutes, I can literally see what the entirety of the world is thinking about.”

Marily Nika

“Create two agents, one that is pro smart-fridge and one that is against smart-fridge… and give me the minimum set of features I would need in order to convince the against agent.”

Marily Nika

“You did your market research in three minutes, and your PRD generation in 90 seconds.”

Claire Vo

“If anything, PMs that use AI are the ones that are gonna take over the role of people who don't use AI.”

Marily Nika

“The best thing you can do is just kill that instance and start over… use gen AI to help you write the best prompt.”

Marily Nika

Questions Answered in This Episode

What specific “minimum convincing feature set” did the pro/contra debate converge on for privacy-first smart fridges, and what did it drop as non-essential?

Marily Nika walks through how she “tool-hops” across AI products to compress a traditional product cycle into ~15–20 minutes: mine user sentiment, synthesize requirements, generate a PRD, build a clickable prototype, and produce a promo video for stakeholder influence.

Get the full analysis with uListen AI

How do you validate that Reddit-mined sentiment is representative enough for early-stage discovery (e.g., weighting by persona, geography, income, or tech-savviness)?

She uses Perplexity’s Reddit/discussions search plus a pro/contra agent debate to surface adoption barriers and the minimum feature set needed to reach product-market fit for a smart-fridge concept.

Get the full analysis with uListen AI

What does Marily’s PRD template include that materially improves PRD-to-prototype results in v0 (sections, constraints, success metrics, non-goals)?

She then turns those features into a structured PRD via a custom ChatGPT PRD generator and feeds the PRD into v0 to create an interactive UI prototype tailored to the requirements (e. ...

Get the full analysis with uListen AI

In a real product review, how would you quantify the impact of bringing a clickable prototype vs. a deck—cycle time, decision rate, or resource commitments?

Finally, she shows how video generation (Google Flow/Veo and OpenAI Sora) can communicate the product narrative, and how NotebookLM can act as a surprisingly effective multi-file “AI judge” for selecting winners at a demo day.

Get the full analysis with uListen AI

What prompting patterns reduce “weird” video-gen artifacts (extra screens, character swaps), and how would you enforce continuity and brand constraints?

Get the full analysis with uListen AI

Transcript Preview

Marily Nika

when all these tools we use every day started coming up, I started figuring out, how can I be an AI-enhanced PM? So how can I be better at my job, have more impact, be more productive?

Claire Vo

So what you're gonna walk us through is, this has inspired some ideas of, "Wow, if I had a fridge, and a video, and data, what product could I build that would be useful to me as a consumer?" Where would you start as a PM who is working on a smart fridge?

Marily Nika

Within minutes, I can literally see what the entirety of the world is thinking about. We have opinions in our hands, and we can split and filter out these opinions based on people that want this versus people that don't want this, and we can actually read and have them debate with each other so that we know what it would take to find product-market fit.

Claire Vo

Okay, you did [chuckles] your market research in three minutes, and your PRD generation in 90 seconds.

Marily Nika

Now, my job is not done as a PM. The next step is to actually create a prototype so that the people that I will present this idea to will see my vision having flesh. [upbeat music]

Claire Vo

Welcome back to How I AI. I'm Claire Vo, product leader and AI obsessive, here on a mission to help you build better with these new tools. Today, we have Marily Nika, a AI Product Lead at Google, and AI educator, who's thinking a lot about how AI is changing the art of product management. She's gonna speed-run us through using some of her favorite tools to do market research, build out requirements documents, prototype complicated things, and a few tricks on how to use AI video gen to influence your stakeholders and sell your vision. Let's get to it. This episode is brought to you by WorkOS. AI has already changed how we work. Tools are helping teams write better code, analyze customer data, and even handle support tickets automatically. But there's a catch: these tools only work well when they have deep access to company systems. Your copilot needs to see your entire code base. Your chatbot needs to search across internal docs. And for enterprise buyers, that raises serious security concerns. That's why these apps face intense IT scrutiny from day one. To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features. Building all that from scratch, it's a massive lift. That's where WorkOS comes in. WorkOS gives you drop-in APIs for enterprise features, so your app can become enterprise-ready and scale upmarket faster. Think of it like Stripe for enterprise features. OpenAI, Perplexity, and Cursor are already using WorkOS to move faster and meet enterprise demands. Join them and hundreds of other industry leaders at workos.com. Start building today. Well, welcome to How I AI. I am excited because we have a very fun, kind of unique product use case you've been thinking about, and using AI to explore the space of what it means to both be an AI PM and how you can PM AI products. So let's dive in, and talk to us about what's been on your mind in terms of products.

Install uListen to search the full transcript and get AI-powered insights

Get Full Transcript

Get more from every podcast

AI summaries, searchable transcripts, and fact-checking. Free forever.

Add to Chrome