Amazon Connect Customer now supports information extraction for agent voice and chat conversations

Amazon Connect Customer has introduced information extraction capabilities that automatically pull key data such as account numbers, reservation IDs, and product names from voice and chat interactions. The feature also derives insights like contact reasons, resolutions, and next steps from conversations. Users can define custom rules to specify what data to extract and when, ensuring flexibility in data capture. Extracted values are available to agents during After Contact Work, supervisors for contact review, and developers via APIs and output files, while sensitive data remains redacted in recordings and transcripts.

The extracted data can be used to trigger automated actions such as email notifications, task creation, and case generation, transforming unstructured conversations into structured workflows. For example, a travel company could automatically extract hotel names, reservation IDs, and call reasons, then populate outbound emails and create follow-up tasks without manual intervention. This reduces handle time and eliminates repetitive data entry tasks for agents.

The feature operates on raw contact content before redaction, allowing specific data points to be captured while maintaining compliance with privacy requirements. Extracted values are accessible through Amazon Connect Customer’s After Contact Work interface, supervisor tools, and developer APIs, including Kinesis Data Streams and S3 output files. This ensures seamless integration with existing workflows and systems.

To implement information extraction, users can refer to the Amazon Connect Customer Administrator Guide for setup instructions and configuration details. The feature’s availability varies by AWS Region, with a full list of supported regions available in the AWS documentation. Additional conversational analytics capabilities are also outlined in the guide.