Marily Nika: PMs who use AI will replace those who don't

Marily Nika is AI product lead at Google and founder of the AI Product Academy, which has trained over 30,000 product managers. This episode was recorded for Lenny Rachitsky’s “How I AI” podcast and published on December 1, 2025. It focuses on one specific question: what does a modern PM toolkit look like when AI is available at every stage of product work?

What the video covers

Nika works through a full product development cycle in real time, using a smart-fridge concept as the demonstration subject. Starting from nothing — no brief, no existing research — she completes the full journey from market research to a polished stakeholder presentation in under 20 minutes. The video is structured around five tools, each applied to a specific stage of the workflow.

Key takeaways

1. Reddit as a research source via Perplexity. Nika uses Perplexity’s “discussions and opinions” filter to surface Reddit debates on the smart-fridge concept, then feeds the output into an AI agent that runs simulated pro/con debates. She finds this more revealing than traditional survey data because Reddit discussions capture authentic friction rather than sanitized responses. The technique transfers to any consumer category with an active Reddit community.

2. Custom GPTs for PRD drafts. Research findings go into a custom GPT configured to match her writing style, which produces a structured first-draft PRD. The output still requires editing, but the starting point is a document with the right sections and framing rather than a blank page. She emphasizes that the GPT’s value is speed to first draft, not quality of final output — the judgment layer remains with the PM.

3. v0 for interactive prototypes. The PRD feeds into v0.dev, which generates a clickable prototype from the document. Nika argues that stakeholders respond differently to interactive prototypes than to written specs: they test assumptions, find gaps, and engage with edge cases they would have overlooked in a document review. A session that would have required a designer and several iterations now takes an afternoon.

4. Flow and Sora for product narrative videos. To create stakeholder communication materials, she generates short product demonstration videos using Google Labs’ Flow and Sora. These compress a product concept into a two-minute visual story that communicates intent faster than a slide deck for audiences who won’t read a PRD. The technique is useful for executive reviews and early investor presentations.

5. NotebookLM as a structured evaluator. At the end of the workflow, she feeds the full set of materials into NotebookLM and uses it as an interactive judge — asking it to evaluate the product concept against market alternatives and surface the weakest assumptions. This functions as a pre-mortem that takes minutes rather than a workshop session.

Worth watching if

You are a PM working without a dedicated designer or research team, since the workflow reduces dependency on specialists at the discovery and prototyping stages. Also useful if you rely on a single general-purpose AI assistant for everything and want to see how chaining specialized tools changes the output quality and speed of each stage.