Dive Brief:
- Amazon Connect announced 29 new agentic AI capabilities, including fully autonomous AI agents and support for human representatives, at Amazon Web Service’s re:Invent conference on Sunday.
- The autonomous AI agents are capable of handling tasks across multiple channels, including voice and chat, according to AWS. They can understand complex requests with multiple intents and take autonomous action on behalf of customers.
- The new agentic capabilities include analyzing conversations, suggesting next steps to live agents and handling tasks like documentation, according to AWS. The support agents can parse conversation context, anticipate needs and proactively offer relevant information.
Dive Insight:
The ability to seamlessly swap between AI agents and human representatives is a core part of Amazon Connect’s offering, according to Pasquale DeMaio, VP of Amazon Connect.
The handoff from AI to human can be initiated at a customer’s request or by the AI agent itself if it doesn’t have the capabilities or permission to handle an inquiry, according to DeMaio. From there, the AI agent can offer context to what has already happened and help guide the call to completion.
“The agentic stuff doesn't stop working,” DeMaio told CX Dive. “It helps the human agent understand what to do. It provides them workflows. It provides them answers. It can even suggest other things to do.”
For example, a customer might call up a company to change the address on file, according to DeMaio. Agentic AI can understand the request, ask the necessary questions and handle the task. However, if the customer is uncomfortable or if a problem that needs human attention arises, the call can seamlessly pass to a live agent.
Companies can monitor Amazon Connect’s AI agents the same way they would a human employee, according to AWS. Supervisors can evaluate AI interactions for accuracy, compliance and customer satisfaction to ensure consistent quality standards.
“We're going to give you the tools to see what the outcomes look like and to look at that day's outcome,” DiMaio said. Contact centers can look at customer behaviors, from individual interactions to long-term data, to help them refine the agentic AI’s behavior.
