Skip to content

Product & UX

This part follows the human-agent relationship from deciding whether a task should be delegated, through control and execution, to recovery and continued work. Architecture owns the underlying mechanisms; this part owns how capabilities, limits, evidence, state, and control are made understandable and usable by people.

  • Product Promise and Task Fit — defines the user, intended outcome, supported and unsupported work, capability boundary, and consequences of failure before choosing an agentic interaction.
  • Delegation and User Control — designs goal and scope setting, interaction modes, action previews, approvals, reversibility, and the user’s continuing authority over delegated work.
  • Progress, Evidence, and Trust — communicates current state, material actions, sources, assumptions, uncertainty, and limitations so that trust follows available evidence rather than fluent presentation.
  • Intervention, Correction, and Recovery — supports clarification, editing, rejection, interruption, takeover, retry, undo, partial results, and recovery when execution goes wrong.
  • Workspaces, Continuity, and Collaboration — preserves tasks, artifacts, history, ownership, notifications, and controllable memory across long-running work, sessions, devices, and people.