About
I build multi-agent AI systems at work and at home.
I am Daniel Chu, a Partner Group Product Manager at Microsoft leading M365 Copilot Mobile product work. Outside the day job, I build and study agent systems because I want to know where they actually help, where they quietly fail, and what product shape makes them safer to use.
Product direction for AI experiences that need to hold up in daily use.
Workflows that move from intent to implementation, review, and recovery.
Making the tradeoffs visible enough that more people can contribute safely.
What I am usually trying to make visible
The parts of AI products that interest me most are rarely the demo moment. They are the follow-through: which files changed, which tool was trusted, what review needs to happen, what happens if the system is wrong, and how a person can recover without starting over.
That is why a lot of my writing circles around agents, evals, workflow design, and product judgment. A useful AI system should not only sound smart in the chat window. It should make the work easier to inspect, improve, and trust.
A production system for making mobile development more accessible while keeping review, evals, and recovery in the loop.
Personal system AtlasA private AI workflow system and public-safe testbed for studying tool calls, sensitive data boundaries, and agent reliability.
A few signals from my background
- As a Partner Group Product Manager at Microsoft, my current center of gravity is Copilot Mobile. I work close to the product surface where AI has to become an actual workflow, not only a clever answer.
- The Copilot Mobile agent system had a concrete adoption signal. PM and Design teammates used it directly to merge a meaningful volume of production PRs with engineering review, enough that review capacity became the next bottleneck.
- I came up through product, startups, and international teams. My background includes Booth, a GNVC Asia win, work across North America and China, and languages that keep me close to different markets.
- The pattern across my work is product judgment plus motion. The recommendations on my profile describe product instincts, bias for action, data-driven thinking, and a collaborative way of leading.
How I work
I like building close to the actual workflow. If the product changes files, the eval should care about files. If the system can write data, the interface should include undo, audit, or repair. If more people can create changes, the review path has to become part of the product.
My current read is that good AI product work is a mix of engineering discipline and taste: constrain the agent where reliability matters, leave room for judgment where the work is ambiguous, and make the cost of mistakes visible before they become expensive.