What changed
Agent assist in Amazon Connect Customer can now respond with interactive components, including cards, carousels, buttons and input fields, instead of plain text only. The AI agent can respond proactively or when the agent asks, and uses views you define first; if allowed in a custom AI agent, it can compose its own card, summary or short form. Agents can select options, fill in details and confirm actions that the AI agent then runs through MCP tools in Connect or third-party services. Responses can also open step-by-step guides, cases, customer profiles and third-party applications with one click. Each interactive response is checked before display, and responses with unsafe links or actions the AI agent isn't allowed to use are rejected.
Vendor documentation (opens in a new tab)Why it matters
This moves agent assist from suggesting answers to completing tasks, such as listing a customer's recent transactions and filing the dispute the agent selects. That can cut handle time, but it also puts real actions behind a click, so the tools exposed to the AI agent and the views it may compose need the same review as any agent-facing integration.
CxCave practical context
Who should care
- Architects
- Administrators
- Developers
- Agents
Implementation considerations
REVIEWCheck the prerequisites: AI agents with a domain and knowledge base, a Connect assistant block in the routing flows, real-time conversational analytics for proactive voice responses, and message streaming for chat. Then pilot the AgentAssistanceOrchestrator agent on one flow with a short list of allowed tools.
CxCave recommendation, not a vendor mandate unless explicitly stated.
