Amazon Connect introduces new metrics to enhance AI agent performance

Amazon Connect has expanded its capabilities by introducing eight new metrics designed to enhance the performance of AI agents. These metrics include the goal success rate, faithfulness score, and tool selection accuracy, among others. The purpose of these metrics is to provide greater insight into the quality of AI-driven customer interactions, allowing for the measurement and continuous improvement of AI agent outcomes.

With these new tools, users can monitor whether AI agents are effectively resolving customer requests, assess the accuracy and reliability of the information provided, and identify any instances of contextual inaccuracies or ‘hallucinations.’ Additionally, the metrics allow for the evaluation of how accurately and effectively tools are selected and used. Customer feedback can also be collected through a simple thumbs up/down rating system when this feature is enabled.

These metrics can be accessed via the AI Agent Performance dashboard in Amazon Connect, or through the GetMetricDataV2 API and a zero-ETL data lake, facilitating custom reporting or integration with existing analytics workflows. The feature is available across all AWS Regions that support Amazon Connect AI Agents. For further details, users can refer to the Amazon Connect Administrator Guide. To explore more about Amazon Connect, which aims to transform every customer interaction into a memorable experience, visit the Amazon Connect website.