Peter Deng on what matters in product and AI
Peter Deng is the architect behind some of the world’s biggest products, with a track record at Instagram, Facebook, Uber and OpenAI.
Peter Deng on what really matters in product and AI
In this interview, he shares everything he’s learned from building products used by billions. What makes him stand out isn’t just what he built, it’s how he thinks about teams, language, systems and human behaviour.
“At Uber, I learned that the product wasn’t the app. It was the price and ETA.”
Peter learned the hard way that what we think of as “the product”, often isn’t what customers care most about. What matters is the whole experience. Uber’s value wasn’t its UX, it was how fast and affordable the ride was.
For companies, the challenge is to balance craftsmanship with the fundamentals of the service being provided. This is especially true for B2B products, where usability can be excellent, but the ultimate priority is the impact on the business.
“A lot of the most successful companies weren’t built on a tech breakthrough. They were built on elbow grease.”
Many of the most successful companies were built not on technological breakthroughs, but on hard work and determination. Peter, with his background at OpenAI, says that you don’t need cutting-edge technology to make a significant impact. Look at Instagram, Facebook, and Uber, they didn’t launch with a massive tech advantage. They connected dots, understood human needs, and built great experiences.
“You have to plan your chess moves in advance. Sometimes, you need to go slow to go fast.”
Going from 1 to 100 isn’t about growth hacks, it’s about systems. Whether it was Newsfeed at Facebook, Messenger or pickups at Uber, the most scalable features were built with structure in mind.
He gives an example from Uber, how the most boring problem of airport pickups became a massive enabler of growth. A whole team worked on just that. Creating the right abstractions made it scale globally.
“Don’t just hire an analytics team, hire a growth leader. They ask the questions that make everyone better.”
Peter sees growth teams as an essential phase for any product that’s scaling. A great growth PM forces you to ask the right questions, get the right data and act on it. This is a mindset shift from guessing to testing.
The point he’s trying to make is that you don’t need to invent new tech to build something valuable, you just need to deeply understand the problem, create a great experience, and build the right systems to scale. The real magic often comes from execution, not innovation.
“You want someone focused on growth sitting next to someone obsessed with craft. That tension is what makes magic.”
One of the best parts of the interview was Peter’s view on team structure. He hires people who see things differently, care about different things and solve problems in different ways. And he sees the leader’s job as holding that tension together.
You don’t want everyone chasing the same thing. You want growth pulling one way, design pulling another, infrastructure pushing from the bottom, research pulling from the top, so the product gets stronger in the middle.
The 5 PM’s archetypes
Peter created a framework for Product Managers hiring that’s both simple and powerful:
Consumer PMs: Obsessed with creating exceptional user experiences, focusing on every detail to ensure delight.
Growth PMs: Data-driven, always eager to experiment, and questioning assumptions in search of solid evidence.
Business PMs: Think in margins and models, aligning strategies for success.
Platform PMs: Build complex systems and easy-to-use tools for other teams.
Delivery PMs: Understand the tech deeply, especially in AI.
Language is a superpower
“Language shapes how we think. And the wrong word in a Product Requirements Document (PRD) can have downstream effects that derail your whole team.”
Peter’s saying that language isn’t just a way to communicate, it’s how we think. The words we choose can guide, confuse or completely change how a team understands a product. He’s seen firsthand how one sloppy word in a PRD can lead to wasted sprints, misaligned teams or the wrong thing getting built.
Think about the difference between writing:
“The system should allow users to manage permissions”
“The system should let admins assign and revoke user roles”
The first is vague: who are the users? What does manage mean? Does that include viewing, editing and deleting? The word manage can send engineering down a different path than what the product intended. The second version is clear, with specific roles and actions, and less room for misinterpretation.
It’s no coincidence that today’s most powerful AI systems are built on language. Language is the condensation of human thought, and that’s why LLMs work.
“People think it’s all about the model. But it’s really about how it fits into someone’s life.”
Peter says the future of AI products won’t be about who has the smartest model. It’ll be about two things:
Access to unique data that makes the AI more useful over time.
Workflow integration that makes it feel like a natural extension of how people already work.
For example: ClickUp vs. ChatGPT. ClickUp didn’t build its model. ClickUp Brain runs on ChatGPT-4o under the hood. What they did well was integrate AI directly into the workflow, including tasks, documents, notes, tickets, and more, right where people already spend hours working. Users don’t have to open a new tool or change context. The AI shows up where they need it, in the middle of their writing process. It feels native. Because of this, many people started using ClickUp Brain instead of jumping between ChatGPT and their notes, even though ChatGPT is technically “smarter.”
That’s why some smaller startups beat established companies. They built something so specific and useful that people made the switch.
Watch the interview
Peter Deng’s advice is thoughtful and battle-tested in some of the toughest environments in the tech industry. If you’re building anything, leading a team or trying to make sense of where AI is headed, this interview is packed with insights worth hearing.