June 2025 Summaries

IRIS Insights, previously known as EarningsEdge AI, is an AI-driven conversational intelligence platform that leverages Hume's Expression Measurement API to infuse emotional insight into language analysis for various organizations, including contact centers and correctional departments. This integration allows for the precise detection of emotional nuances such as frustration and confusion, enhancing customer and inmate interactions by improving metrics like CSAT and etiquette scores and reducing indicators of low engagement like dead air. Traditional sentiment tools were inadequate for IRIS's needs as they lacked the granularity required to capture subtle emotional shifts, leading IRIS to adopt Hume's API, which provides detailed emotional data with 48 distinct emotions per utterance. This enables proactive intervention and personalized coaching, resulting in improved agent performance and offering a competitive edge over conventional sentiment analysis tools. IRIS's emphasis on emotional intelligence has not only enhanced client visibility into customer sentiment but also led to regulatory approval for piloting its correctional platform, underscoring the significance of emotional intelligence in public safety.

The Hume AI MCP Server allows users to interact with the Hume AI assistant directly through client applications that implement the Model Context Protocol (MCP). This enables tasks such as narrating audiobooks or designing voice characters for video games. The server's development revealed issues with providing a quality experience, including failure to use the right voice, format text correctly, and continue audio from previous generations. To address this, the team put on "prompt engineer hats" and began evaluating the prompt descriptions and field descriptions. They realized that building a production-grade experience required more comprehensive evaluations. The problem of capturing long interactions proved challenging, with existing literature primarily focusing on traditional question/answer type prompts. The team explored different approaches to evaluate the MCP server, including single-response, postponement-tolerant, manually-extended evals and role-play-based evals. The latter approach, which involves offloading task extension to LLMs, was chosen as the most direct method for evaluating the server's performance. This approach allows for a more straightforward evaluation process by analyzing entire transcripts rather than breaking them down into individual stages. While there are weaknesses to this approach, such as tooling limitations and high cost, it provides a practical indicator of whether changes to the prompt improve assistant behavior. The MCP server and its evals are open source, making this approach accessible to others.