AI and product management | Marily Nika (Meta, Google)

Marily Nika opens the episode with a warning she says she gives people repeatedly: beware the "shiny object trap". Her rule is "don't do AI for the sake of doing AI." A team should first confirm that a real pain point exists that deserves a smart solution, sketch the solution only at a very high level, and only then work out how to implement it.

Lenny introduces Nika as the teacher of a popular Maven course on AI and product management and as a product lead at Meta working on the Metaverse, avatars and identity. He says she previously spent over eight years at Google on Google Glass, computer vision and machine learning for speech recognition. He previews the conversation: what PMs should watch in AI, resources for getting started, how today's AI tools can already help a PM, and a relatively technical look at what a model is and how models are trained.

The recording also contains two sponsor reads, one for Amplitude (analytics) and one for EPPO (an A/B testing platform), delivered before the interview proper begins. They are advertising and carry no claims from the guest.

Lenny frames the first problem as one of pace: as a former full-time PM he finds it hard to keep up with product and technology change, and it feels harder in AI because something seems to come out every day. He asks which media Nika uses to stay on top of AI and machine learning.

Nika's first answer is newsletters. She subscribes to Lenny's and names The Download from MIT's technology publication (the transcript renders it as "MIT Ecology Review") and another called Atelier. She notes that these are not necessarily AI-centric, which fits her broader claim that in the future everything will be AI by default, so even technology-focused reading will have AI "sprinkled in".

Asked what is overhyped and what is underhyped, Nika says the topic is both at once. She cites an article she read that morning in which writers fear that online writing will die and their jobs will be taken. Her view is the opposite: ChatGPT and similar technology are "enhancing our work" rather than stealing from people. As an under-hyped example she mentions a research article she read saying AI can now detect lies, which she says makes lie detection possible for security or workplace uses. This is her second-hand account of one article, and the transcript gives no further detail on the study or its accuracy.

Her practical advice is to read widely in newsletters and online blogs, because AI is not only about chat. The recommendation is general and the transcript names no other specific sources beyond those above.

Nika describes three concrete uses of ChatGPT at work. First, mission statements: she pastes in her own version and asks for a rewrite, and says even the first attempt is often fantastic. Second, user segments: it surfaces segments she says her own mind would not reach, along with their motivations and pain points. Third, ideas, including ideas for AI-enhanced features. The transcript is garbled around the last point, so its exact meaning is uncertain.

She is explicit about the limit of this use: "I'm not making it do my job for me." She asks the tool only after she already has a mission in mind and knows what she wants to do. Lenny confirms the workflow, which is to give it her draft and ask for a better one, and she agrees.

Her reasoning for why the rewrite is better is about audience. A mission statement is read by leadership, junior staff, stakeholders, other departments and competitors, not only by PMs who already have context, so it must use words everyone can understand. She says a kid could understand them and still be inspired. This is her judgment of the output, offered without a comparison or measurement.

For personas, she gives an example prompt: if you are working on a fitness band with no screen, you ask who would be interested in it. The answer she describes is a bulleted list, such as young professionals who are interested but short on time, and people who do not want to charge their wearables every day.