Figma: designing agentic tools that keep users focused on their goal
When AI agents take actions — scheduling, summarizing, routing, executing — the design decisions that determine whether people trust them are not primarily visual. They are structural. Sheta Chatterjee, Head of UX and Design at Google Cloud AI, published a piece on the Figma blog in April 2026 drawing on enterprise deployments where agents operate in high-stakes environments.
The core principle is stated directly: “The user’s focus should always be on their goal, not on managing AI.” This sounds obvious until you examine how many enterprise AI tools invert it, requiring users to understand model limitations, repair broken outputs, and track what the agent has done across multiple sessions. Chatterjee treats that friction as a design failure, not an AI limitation.
Transparency as a trust mechanism takes several concrete forms in the described approach. The agent narrates its intended actions before executing them, giving users a checkpoint to intervene. Source citations appear in all responses, showing where information came from. An “AI Inbox Dashboard” provides a visual overview of agent tasks, their completion status, and which items require human review — turning the agent’s behavior from a background process into a visible team function.
Governance is embedded at the design level rather than treated as a separate administrative layer. An “Agent Designer” feature lets users define which data the agent can access and when human approval is required before it proceeds. For enterprise contexts where audit requirements are real, this makes scope definition part of the product rather than a policy document attached to it.
The shift from individual chat threads to shared project spaces is another structural choice. When an agent operates inside a shared workspace rather than per-user sessions, its outputs are visible to the team, its decisions become accountable, and information does not get siloed in one person’s conversation history.
Chatterjee’s closing point addresses a tendency in AI product development: craft and intentional design decisions matter more than rapid prototyping when the goal is user trust. Enterprise users abandon an agent that makes them feel out of control faster than one that is capable but transparent.
Useful for product designers and PMs building agentic products for professional environments where accountability and auditability are requirements, not optional features.