ReleaseBytes
aws AWS Contact Center Blog ·

AWS Showcases Voice-Activated AI Agent for Amazon Connect Supervisors

blogaiawsgaarchitectaws-bedrock
announcement

AWS has published a new solution for creating a voice-activated AI agent that empowers Amazon Connect supervisors with real-time operational control. This agent allows supervisors to manage production AI agents conversationally via phone, enabling rapid changes to prompts or intents without requiring console access. It addresses critical operational agility challenges in AI-powered contact centers, especially during service outages or when immediate agent behavior adjustments are needed. Built with Amazon Connect Customer, Amazon Bedrock, and the Model Context Protocol, the solution offers secure, PIN-authenticated access and is available as a GitHub repository.

  • →Addressing Contact Center Operational Agility Challenges
  • →Voice-Activated Supervisor AI Agent Solution Overview
  • →Secure Voice Access and Two-Factor Authentication
  • →Conversational AI Orchestration with Amazon Bedrock
  • →Model Context Protocol (MCP) Tool Execution
Notes (5) ›
  • Addressing Contact Center Operational Agility Challenges

    The solution targets the challenge of rapidly adjusting AI agent behavior in Amazon Connect Customer, particularly during system outages or when specific intents need to be toggled. Traditional methods involve multiple console logins and manual configuration, leading to delayed responses and degraded customer experience.

  • Voice-Activated Supervisor AI Agent Solution Overview

    AWS presents a secure voice-activated AI agent for supervisors to manage production AI agents conversationally via phone. This eliminates the need for console access and allows immediate changes to payment intents or agent prompts, with effects taking place in seconds.

  • Secure Voice Access and Two-Factor Authentication

    The solution implements defense-in-depth authentication using a phone allow list and PIN verification via AWS Lambda and Secrets Manager. It also incorporates brute-force protection with an Amazon DynamoDB table, locking out callers after three failed PIN attempts.

  • Conversational AI Orchestration with Amazon Bedrock

    A Bedrock-powered AI agent handles the conversational interface, guided by a carefully crafted system prompt. This prompt mandates PIN authentication, defines the supervisor persona, lists available tools, requires confirmation for destructive operations, and instructs the agent to report test results.

  • Model Context Protocol (MCP) Tool Execution

    The AI agent invokes nine Model Context Protocol (MCP) tools, built as Amazon Connect Customer flow modules, to perform real-time operations. These tools interact with backend Lambda functions to manage production AI agent capabilities, prompts, and intent configurations.

Read the original announcement →

https://aws.amazon.com/blogs/contact-center/voice-activated-ai-agent-with-amazon-connect-customer-and-amazon-bedrock/

Related releases