The Role of a Product Manager in Embodied Intelligence
Published:
Originally published on Substack.
Three months ago, I stepped away from my startups, closed all my past projects, and decided to work for a company.
Back in April, I walked away from a $7M robotics deal because the technology simply wasn’t ready for mining exploration—it still isn’t even today. However, that project gave me valuable insight into embodied AI, which led me to join an embodied AI startup.
In the weeks after joining, I quickly realized that the role of a product manager in embodied AI is very different from a typical PM who works only in software or hardware. It’s a huge challenge—and one worth sharing. So I wrote this short article to illustrate what makes a good PM vs. a bad PM in embodied AI. I’d like to share it with all of you.
The Role of a Product Manager in Embodied Intelligence
In the era of embodied intelligence, a product manager’s role goes far beyond writing requirements or aligning schedules. A truly great PM can bridge AI, hardware, software, and real-world deployment.
To create embodied robots that work reliably in the real world, a PM must be able to:
Guide the team with a clear vision
Stay decisive in complex hardware-software collaboration
Turn real customer pain points into engineering and product priorities
Drive results even in uncertain environments
This guide—“Good PM / Bad PM”—aims to align the team on the responsibilities, value, and boundaries of this role.
We also encourage the team to:
Hold PMs accountable—speak up when necessary
Point out when PMs are not demonstrating the “good” behaviors
Support and trust PMs when they deliver
Remember: a great product is always a team effort.
Good PM vs. Bad PM
Vision & Mission Alignment
Good PM: Makes decisions around the core mission (real-world deployment), translates strategic vision into concrete goals, focuses on deployment over flashy demos.
Bad PM: Chases trends, lacks strategic judgment, substitutes slogans for action plans.
Execution & Accountability
Good PM: Takes responsibility for results, not just process; aligns the team with clear, verifiable PRDs; distinguishes between demo and mass production.
Bad PM: Shifts blame, provides vague documentation, fails to drive cross-team problem-solving.
Customer & Market Understanding
Good PM: Understands real-world scenarios (warehouse, factory, retail, hospital), identifies real pain points (reliability, safety, efficiency), and validates continuously in the field.
Bad PM: Works in isolation to impress investors, ignores customer feedback, focuses on flashy tech instead of real needs.
Cross-Team Collaboration
Good PM: Speaks both engineering and business languages, proactively builds shared understanding, turns disagreements into consensus.
Bad PM: Pushes work onto teams, thinks sending documents equals collaboration, lacks alignment mechanisms.
Prioritization & Focus
Good PM: Focuses on high-value problems, dares to say “no,” delays low-priority projects, emphasizes impact over quantity.
Bad PM: Says yes to everything, confuses priorities, spreads resources thin, producing little.
Communication & Clarity
Good PM: Keeps information transparent, explains complex technical and business trade-offs clearly, ensures everyone understands why.
Bad PM: Hides behind jargon or slides, communicates results without reasoning, is passive in communication.
Data & Evidence
Good PM: Uses core metrics (stability, task success rate, MTBF) to drive iterations, actively collects and validates data, and uses data to tell a story.
Bad PM: Acts on gut feeling, doesn’t track KPIs, believes AI will “improve itself.”
Speed & Resilience
Good PM: Moves fast without sacrificing quality, removes obstacles daily, and has backup plans for hardware challenges.
Bad PM: Waits for perfection, delays addressing problems, underestimates the complexity of software-hardware integration.
Leadership & Culture
Good PM: Energizes the team, builds influence through trust, transparency, and results, and upholds product integrity.
Bad PM: Plays politics, shifts blame, avoids responsibility, pursues personal credit.
Long-Term Thinking
Good PM: Designs systems, not temporary features; adopts a platform perspective (SDKs, data flywheels, services); balances current delivery with future scalability.
Bad PM: Focuses only on short-term KPIs, accumulates technical debt, treats the roadmap as a wish list.
A great product is built by a team that shares vision, trust, and accountability.