Palantir: Design Patterns for Human-Agent Collaboration
Palantir designers share design patterns for legible, trustworthy human-agent collaboration: presenting agent reasoning without overload, with clear decision attribution.
"Every AI product seems to look the same. Yours doesn't have to." That's how Palantir opens this design deep-dive from DevCon 6. In about 17 minutes, product designers Emily Su, Lide Li, and Philipp Shchekin walk through how they design the collaboration surface between humans and AI agents, aiming for three qualities: seamless, legible, and trustworthy.
Two core problems run through the talk: how to present an agent's reasoning so users understand why it acted without drowning them in information, and how to build clear attribution so users always know whether a decision was made by a human or by an agent. Their approach is to design these collaboration patterns and then encode them into Palantir's Ontology — making the patterns part of the system itself rather than scattered UI conventions.
In a landscape of chat-box-shaped AI products, this is one of the few first-hand accounts of what agent UI should actually look like. Best for designers and PMs building agent products — especially teams whose products have both humans and agents making decisions, and who owe users a clear answer to "who did this?"
FAQ
What is this video about?
A design deep-dive from Palantir's DevCon 6 on design patterns for human-agent collaboration — making it seamless, legible, and trustworthy. The talk runs about 17 minutes.
Who are the speakers?
Three Palantir product designers — Emily Su, Lide Li, and Philipp Shchekin — presenting at DevCon 6, published on Palantir's official YouTube channel.
What core design problems does the talk address?
Two: how to present agent reasoning so users understand its behavior without information overload, and how to build clear attribution so users always know whether a decision came from a human or an agent.
What role does the Ontology play?
Rather than leaving collaboration patterns as UI conventions, the designers leverage Palantir's Ontology to design and encode them — making human-agent collaboration a system-level capability.
Who should watch this?
Designers and PMs building agent products — especially teams where humans and agents both make decisions and users need to know who did what. You don't need Palantir experience to apply the design thinking.