Artificial intelligence has moved quickly from experiment to default layer inside collaboration software. Meeting tools summarize conversations. Coding environments suggest the next edit. Enterprise platforms promise assistants that can draft, search, and decide alongside people at work. The investment behind those features is enormous. The user experience often still is not.
Too many AI products still feel bolted onto existing workflows rather than designed into them. Capability arrives first. Clarity arrives later, if at all. Users are left to reverse-engineer what the system can do, what it might get wrong, and whether they should trust it with real work.
That gap (between technical power and everyday usability) is where product designer Peng Zheng has spent much of his career. Trained in human-computer interaction and shaped by product organizations at companies including Zoom, Bloomberg, IDEO, and frog design, he approaches AI product work as systems design: ordering attention, surfacing uncertainty, and making complex software feel legible enough to adopt.
In an industry racing to ship intelligence, his through-line is quieter and harder. Great AI collaboration products are not defined by how impressive the model is. They are defined by whether people can understand the system, enter it confidently, and keep using it every day.
The Trust Problem in AI Collaboration Software
In traditional software, design often answered a familiar set of questions: Where do I click? What happens next? How do I recover if I make a mistake?
AI collaboration products add a second layer. Users are no longer only navigating screens. They are negotiating with systems that generate, recommend, and act. The questions become sharper: Can I trust what this is doing? When should I intervene? How do I know the system is uncertain?
Peng’s practice sits at that second layer. He treats interface design as attention design: how information is prioritized, how risk is communicated, and how multi-platform workflows stay coherent when intelligence is present. The goal is not to hide complexity behind a chat box. It is to make complexity manageable.
That distinction matters as companies move from AI demos to AI defaults. A feature that looks magical in a launch video can still fail in production if users cannot predict its behavior. Trust, in this sense, is not a marketing claim. It is a design outcome: clear boundaries, readable states, and interaction patterns that keep the human in control while the system assists.
From HCI Training to Enterprise Product Surfaces
Peng’s foundation is human-computer interaction at Carnegie Mellon University, a training ground that emphasizes research, evaluation, and the distance between what teams intend and what users actually experience. That academic grounding shows up less as theory for its own sake and more as a working habit: begin with the user’s decision moments, then ask whether intelligent systems can improve those moments without adding noise.
His professional path then moved through environments where that habit was tested at different scales and densities of information.
At Bloomberg, the work lived inside information-dense professional tools, surfaces where clarity under complexity is not optional. At IDEO and frog design, the emphasis shifted toward research-led, systems-oriented design practice: framing problems carefully, testing assumptions, and treating product experience as an architecture rather than a collection of screens.
Those stops matter because they map the wider shift happening across the industry. Product design is no longer only visual craft. In AI-era software, it has become a translation layer between engineering capability, business goals, and human behavior. Peng’s career tracks that expansion directly.
Zoom: Designing AI Companion and Navigation at Scale
The clearest expression of that approach came at Zoom, where Peng worked at the intersection of AI and everyday collaboration.
As a product designer on AI Companion, he helped grow the feature’s user base from about 2.2 million to 5 million users. In parallel, he led a comprehensive redesign of Zoom’s desktop navigation system, shaping a core experience within a platform that reached more than 300 million daily meeting participants at the height of its global adoption.
Those two projects sit at the hard center of modern product design. One is about introducing intelligence into a product people already depend on. The other is about organizing the product’s primary structure so millions of people can find their way through it without friction. Together, they require the same discipline: scale, clarity, and the quiet work of making powerful software feel approachable.
AI Companion work, in particular, highlights why design becomes decisive once a model leaves the lab. An assistant that can summarize or suggest is only useful if people understand when to invoke it, how to evaluate its output, and how to continue their workflow afterward. Navigation redesign makes a similar point from another angle. Before users can trust an intelligent feature, they have to be able to reach it, and return to the rest of their work, without cognitive tax.
In both cases, the design problem is not decoration. It is infrastructure for attention.
From Cursor to Grok Bot: Designing AI Agents People Can Steer
What distinguishes Peng’s path is continuity rather than reinvention. Across Bloomberg, IDEO, frog, Zoom, Anysphere, and now SpaceX, the companies change, but the question stays stable: how do you turn abstract technical capability into experiences that feel predictable, human, and worth returning to?
At Anysphere, Peng recently worked on Cursor, the AI coding agent, helping shape how developers collaborate with AI inside real software workflows. That setting made the next phase of product design unusually clear. Coding tools are not only adding chat. They are changing the unit of work itself: from typing every line to steering systems that propose, edit, and execute.
Today, he is driving the design of Grok Bot at SpaceX, extending the same design thesis into a broader agent experience. In that environment, the designer’s job expands again. The interface becomes less a static map of screens and more a control surface for intelligent behavior. Users need to see what the system is doing, where it is uncertain, and how to intervene.
Peng’s background in HCI and enterprise collaboration products is well suited to that shift, because the scarce skill is still judgment: deciding what should be simple, what should be explained, and how a complex system earns daily trust.
Recognition That Follows the Product Work
Recognition has followed the product work rather than standing apart from it. Peng’s design practice has received awards including the 2025 IDEA Award, the 2025 Red Dot Award, and the 2026 iF Design Award, along with an Apple App Store Awards 2024 nomination for the Zoom Apple TV app.
In 2026 he also served as a judge for the Lovart AI Design Challenge, and was selected for the inaugural a16z Design Engineer Fellowship, a selective cohort of AI-native design leaders working at the boundary of craft and technical fluency.
Those markers matter in context. They are not the story by themselves. They are supporting evidence for a practice centered on shipping and shaping AI collaboration experiences at real scale. For TechBullion readers evaluating product leaders in the AI era, the more useful signal is the through-line: systems thinking, trust-centered interaction design, and a career spent translating capability into use.
What Product Design Means When Software Becomes Intelligent
Taken together, Peng’s story is not about leaving product design for something else. It is about deepening what product design means when software becomes intelligent.
In an era when AI can generate interfaces quickly, generation is no longer the bottleneck. Judgment is. Teams can produce options faster than they can decide which options deserve to ship. Models can propose actions faster than users can evaluate them. The product designer’s contribution is to hold the system together: to keep intelligence inside a workflow people can understand, enter, and trust.
That is why Peng’s work keeps returning to the same core questions. How should attention be ordered when the product can do more than the user asked for? How should uncertainty be shown without creating fear or noise? How should navigation and interaction patterns scale when millions of people depend on the same surface every day?
The companies on his résumé answer those questions in different product languages: from enterprise collaboration at Zoom, to recent work on Cursor at Anysphere, to Grok Bot at SpaceX. The design thesis underneath them is shared: AI collaboration succeeds when design treats intelligence as a system, not a feature.
For product organizations building the next generation of assistants, copilots, and agentic workflows, that thesis is becoming less optional. Technology will keep moving quickly. The products that last will still be the ones people can understand, trust, and keep using, and that principle does not change.
