What it takes to be a top PM today | Robby Stein (Google Search)
In a Nutshell
The core shift in product management is from execution and coordination to judgment and taste—now that anything can be built, PMs must make excellent decisions about what deserves to be built. The three-chapter framework for building great products is: deeply understanding users' real jobs-to-be-done through storytelling and context, systematically diagnosing and fixing root causes of friction rather than surface symptoms, and obsessively perfecting the final experience so the product works flawlessly and creates joy. AI amplifies each stage—scaling user research, analyzing qualitative feedback for root causes, and automating quality testing—but the fundamental craft remains making great decisions about real people.
These notes were generated by AI and may contain inaccuracies.
Product building has never been as interesting as it is now. After nearly 20 years of building products in the consumer sector, Robby Stein has built startups, helped lead teams through major changes at Instagram and Google Search, and created products used by billions of users while also building products that no one uses. A recurring pattern emerged from successful products versus unsuccessful ones.
At Instagram, Stein's team built and launched Instagram Stories along with related experiences including Reels, Feed arrangement, and direct messaging. Currently at Google Search, his team is working on a new experiment to rethink Google Search in the age of artificial intelligence, enabling users to input anything on their mind and receive the most useful information across videos, audio, and text, leveraging web and Google knowledge.
A large part of product manager value used to lie in getting things done through organization, creating energy and momentum, functioning as both project manager and product manager. In today's world, almost anything imaginable can be built. The true value of a product manager now lies in judging things—it's a matter of taste and accomplishing something perfectly.
The core business of a product manager involves doing many things fairly well, but the craft of product management centers on one essential skill: decision-making. Being educated about what it takes to make great decisions requires humility, as many decisions are wrong.
An unwritten book in Stein's mind consists of three chapters about the process his teams used to build successful products. The first chapter is understanding people deeply—it all starts with people and their basic needs.
Clayton Christensen's book "Competing Against Luck" presents the Jobs to Be Done framework, which Stein's team at Instagram adopted after conducting an external workshop with the framework's creators. The core concept is that people don't use products—they "employ" them to do things for them.
The best way to extract these insights is through storytelling. When talking to users, ask for details about their life circumstances: How was that day? Where were they? What were they reading? Who were they with? What were they wearing? Were they with children? Alone? In a store? Connected to the internet?
Stein's last major purchase was a bed. He and his wife were looking at multiple beds that looked great, but his wife asked him to jump on one and make a circular motion. When asked why, she explained: "The main thing I need is a bed that doesn't wake me up if I move." This insight about motion transfer would never have been discovered through conventional product thinking focused on cooling features, environmental friendliness, or price.
In early AI experiments at Google, the team realized people come for inspiring journeys, but chatbots relied solely on text. They invested in advanced multimedia understanding, retrieval, and knowledge systems so users could ask about a sofa and receive visual responses, enabling multi-round conversations about emerald green color, soft touch, or decorative pillows.
Another early AI model exploration revealed that while people could ask about neighborhoods or food, the experience lacked the knowledge, accuracy, and visual elements people expected from Google—seeing places on maps, star ratings, closing times, food types, and costs. Combining conversational AI with these trusted elements created the magic product that remains one of the most popular features.
AI can transform user research by taking transcript data from interviews to precisely define required tasks and scale the process. A smart agent could interview people using the same methodology, asking users broadly about why they use products and underlying problems.
After gaining initial vision of the problem and understanding what people want, teams create something that usually doesn't feel great initially. Most successful products started off badly. The process involves gaining a clear analytical view of subtle problems, prioritizing them, understanding them deeply, fixing them, and repeatedly asking questions in a continuous loop—essentially conducting a training process similar to model training.
When Instagram Stories launched successfully, not everyone felt comfortable participating despite the premise being an opportunity to share daily life. The team identified the root cause: audience problems. Users worried about ex-partners, teachers, aunts, and sisters seeing their content. A massive quantitative survey with thousands of participants revealed audience concerns as the primary barrier that no creative tool could overcome.
The team spent two years experimenting with ideas including "favorites," backdoor profile access, and posting to feeds under "close friends" tags. These were confusing and didn't work. The only successful approach was within Stories themselves. They deleted the product entirely and relaunched it as a story-based experience.
Reels launched in Brazil with the assumption that silly dances shouldn't remain on profiles permanently. The launch failed because people investing time in great dances didn't want content to disappear. Users wanted to be entertainers building businesses through content that would continue and spread widely.
The team made a comprehensive update: Reels became a standalone permanent format rather than temporary, visible to everyone's school rather than disappearing. This change enabled the trend's great success.
At Google, an "anti-gravity" system collects feedback from users comfortable sharing experiences. AI models can now process this qualitative feedback library. In a backpack shopping example, users noted that while the system knew backpack dimensions, it should have asked about the child's height since the bag would be too heavy otherwise.
The concept of collaboration and creating dialogue to help users progress toward goals proved crucial for user satisfaction. The team invested significant energy in this area, seeing improvements in interaction, usability, value, and benefit.
When products reach the stage of being ready for major launches, preparation involves finishing touches and final product elements—all about perfection. People can tell if creators really care about what they've made.
Two driving factors: whether the product works perfectly without user suffering, and how it makes users feel—whether it brings joy and enthusiasm.
Stein built an internal tool using the "Anti-Gravity" system that uses AI agents to test products by going to Google, typing questions, asking follow-up questions, recording experiences, taking screenshots, and evaluating based on comprehensive criteria.
The model evaluates whether math questions receive proper LaTeX presentations, whether visual topics like bioluminescence include visual displays, and identifies flaws for self-correction. This replaces traditional QA teams and obsessive product usage by team members.
Google reinvented its search box after 20 years, announced at I/O conference. The AI-powered search box grows with users, allowing file, video, and image uploads with context. Color gradients represent AI power behind the product. The indicator flashes through Google's different colors, creating moments people notice and share on X. Clicking inside the search box creates an excited jumping animation with vibrations.
Through movement, colors, texture, touch, and experience intention, products convey humanity and give them soul.
Product managers must still do the same fundamental things, but now possible in completely new ways with AI:
- Understand people deeply and get to core needs and problems
- Be meticulous and analytical in identifying what needs fixing for product success
- Build products that are cool, easy to use, work perfectly, and make people feel great
The craft of product management has never been more important. The skill of making great decisions can lead to great products. Product managers should spend less time thinking about cool features, techniques, or presentation methods, and more time thinking about real people on the other side of the screen.
Keep Lenny's Podcast in your library
Save the videos and channels worth coming back to, and find them again in one place.
