Create, customize, and deploy AI-powered chat agents in Amazon Quick Suite

Generate content and provide answers through natural language conversations

Analyze and summarize information from connected spaces, dashboards, topics, datasets, and uploaded files

Invoke actions to perform predefined steps for consistent, repeatable outcomes

System chat agent – This chat agent ("My assistant") is automatically available to all users by default. The system chat agent serves as a base planner with no inherent data or actions of its own—it dynamically accesses resources available to each user during chat time, allowing it to be tailored to individual user permissions and available content. Admin users control system chat agent settings by assigning specific users as owners so they can customize its persona and other settings.

The system chat agent is enabled with all chat capabilities, including file upload functionality, LLM knowledge access, toxicity and other guardrails, web search. It also includes chat data scoping mechanisms that provide access to spaces, topics, dashboards, knowledge bases, and actions based on user permissions.

Custom chat agents – These chat agents can be created and customized to specific use cases by users with chat agent creation capabilities in Amazon Quick Suite, and shared with anyone. Admins can choose to restrict specific users and groups from creating custom chat agents, while still allowing users to use chat agents via chat. Chat agents return responses scoped to content that their invoking users have permissions to.

Preconfigured with resources – These chat agents use only the configured resources (for example, spaces as knowledge sources, action connectors as tools) when looking for answers or orchestrating actions as their defeault behavior. While chatting, users can attach additional resources as per their need or invoke actions directly. For example, chat agents configured with only spaces containing files can't take actions by default unless users explicitly invoke them. Similarly, chat agents configured with only actions rely on LLM knowledge unless users attach a space or dashboard for enterprise-specific answers. However, users may attach additional spaces or directly invoke other actions they may have access to within the chat interface.

Not opinionated at build time – These chat agents are not initially configured with resources when they are built (spaces, or action connectors, or both). Chat determines the resource boundary for the chat agent. For example, if no space is configured for the chat agent, chat will default to all spaces or LLM knowledge until the user makes a change. If no action connectors are configured for the chat agent, during chat, all actions a user has access to will be available to the agent until user makes a resource selection that doesn't include actions.

User resource selection during chat affects chat agent behavior. If a user selects a space, the chat agent will only answer with data within that space and take actions available within that space (if it is not opinionated at build time). If a user selects a dashboard, topic, or knowledge base, the chat agent will only answer from that source and not take any actions since the focus is changed to a specific data source. Users are expected to select all relevant resources within chat for comprehensive assistance.

Owner – Owners can edit, share, and delete the chat agent.

Viewer – Viewers can view and use the chat agent.

You can completely disable all chat agent functionality for users, including chatting with the default agent, chatting with custom agents, and creating new agents. This can be done by restricting the Chat agent capability.

You can also specifically restrict the ability to create agents without impacting chat with agents. This can be done by creating a custom permission profile and restricting only the feature Create chat agents.