Agent in chat
July 28, 2025
How you place an AI agent shapes usability and decision-making and thus AI product design! 🤖 This pattern places the AI agent’s “thinking,” updates, and intermediate steps directly into the chat stream as if it were another user.
It mimics human conversation and makes the agent feel like a collaborator or creative partner.
Key traits:
Conversational style: Reasoning and status updates appear as messages, making interactions feel fluid and natural—ideal for co-creation, critique, and exploration.
Low cognitive switch-cost: Updates are close to where the user inputs commands, making it easier to follow short turn-by-turn exchanges.
Context risk: As more messages are exchanged, previous agent updates may scroll out of view, making it harder to maintain continuity across complex tasks.
Mental model: Positions the agent as a peer or partner—users treat the system more like a person with shared agency.
Used in: Lovable, Figma Make, Notion AI