Stretto has built an AI voice agent called Tara using Amazon Nova 2 Sonic, a speech-to-speech model available in Amazon Bedrock, according to a case study published by AWS. The system conducts bankruptcy intake interviews over the phone for the company’s Bankruptcy.com site.
Stretto previously used a web form that captured only basic information: debt amount, attorney status and contact information. Bankruptcy attorneys preferred leads with more detailed financial data, including debts, assets, vehicles, real estate, employment history and prior filings. However, extending the web form reduced completion rates, and staffing a human intake team was cost-prohibitive, according to the company.
Tara picks up incoming calls, introduces herself and guides callers through a structured conversation collecting personal information, financial details, asset inventories and employment history. The system adapts its tone to the caller’s emotional state, according to AWS. After the call, Stretto’s system applies attorney-defined rules to sort leads.
Stretto worked with the AWS Generative AI Innovation Center and AWS Partner CI&T to develop the solution. The company cited requirements for low latency, clean interruption handling, emotional awareness and cost-effective unit economics when handling sensitive financial conversations.
The case study covers the architecture and phased build from validation through production deployment, as well as voice quality testing for sensitive conversations.