USAA

USAA, one of the largest financial services providers in the U.S., operates at significant scale — serving millions of military members and their families with auto insurance, banking, and investment products across a workforce of 38,000 employees. Member service representatives (MSRs) faced mounting cognitive load navigating disparate systems to resolve complex, multi-channel inquiries in real time. Simultaneously, software and data engineering teams struggled to meet growing IT demand without proportional headcount increases. Across claims, underwriting, and servicing operations, the inability to efficiently extract signal from massive volumes of unstructured data — documents, audio, images, and text — slowed decisions and increased operational cost.

USAA built a suite of internal Generative AI tools through agile AI pods of 10–12 cross-functional team members, deliberately targeting internal users before any member-facing deployment. The MSR Co-Pilot integrates with existing service workflows to surface relevant information, summarize member interactions, and capture follow-up actions in real time — reducing the cognitive burden on representatives handling complex inquiries. A GenAI pair-programmer was deployed for software and data engineers to accelerate code generation, documentation, and test-data creation. A third system was built to ingest unstructured employee feedback from internal Slack channels at scale, applying GenAI to identify themes and sentiment trends across thousands of daily messages. All three tools were developed using rapid pilot cycles before broader rollout.

The employee feedback analysis tool was designed, built, and deployed in just eight weeks — a fraction of the timeline typical for enterprise AI or traditional software projects of comparable scope. Across the organization, 38,000 employees received AI awareness training, establishing a foundation for responsible adoption at scale. Key outcomes include:

• Feedback analysis system delivered in 8 weeks vs. traditional multi-quarter timelines
• IT demand absorbed through productivity gains rather than headcount growth
• Hundreds of AI solutions deployed in earlier phases, with GenAI now accelerating the next wave

USAA expects GenAI investment returns to significantly exceed costs, with formal productivity and quality metrics being tracked for the pair-programmer rollout.

• Start internal, then expand: Deploying GenAI to employees first — before members — builds the reliability track record needed for higher-stakes use cases in auto insurance claims and underwriting.
• Eight-week pilots are achievable: Agile AI pods of 10–12 cross-functional members can compress enterprise delivery cycles dramatically when scope is tightly defined.
• Governance must scale with deployment: Unstructured data handling (audio, documents, images) requires explicit data governance strategies before GenAI can operate reliably across insurance operations.
• Measure productivity, not just speed: Tracking coding quality alongside velocity ensures productivity tools don't introduce downstream defects.