Ansira builds always-on AI voice support on AWS with Thunder

For a company whose business runs on the connection between enterprise brands and their local partners, every unanswered call carries a cost. Ansira’s central customer service hotline fields questions from thousands of partners across more than 500 brands: franchise owners, channel partners, agents, dealers, retailers, and franchise owners who depend on Ansira to help them run co-op advertising programs, place media orders, and navigate marketing asset sites within the Ansira platform. With live agents available only during set business hours, partners calling after hours, on holidays, or during peak periods had no option but to leave a message and wait.

The pressure intensified after Ansira acquired two competitors in 2024, requiring it to consolidate three separate companies, each with its own support system, onto a unified Salesforce Service Cloud environment. The migration compressed resources across IT and support teams just as call complexity was rising. Tier-one questions like password resets, order status checks, and how-to guidance for marketing tools consumed agent time that would be better spent on nuanced, brand-specific issues. “We really pushed a lot of AI initiatives to gain efficiencies and get more streamlined,” says Christina Klupe, vice president of partner support at Ansira. The question was not whether to act, but how fast.

As Thunder guided Ansira through the Salesforce migration, the team applied the AWS Business Value Realization (BVR) framework to map where an AI investment would deliver the greatest return. That analysis pointed to a clear opportunity: Ansira’s general support line was absorbing high volumes of routine, repeatable inquiries that were consuming agent capacity without adding strategic value. Adding an agentic AI voice layer was the natural next step. “Ansira has been intentional about adopting AI where it can create real value for our business and our clients, and this solution felt like the right fit,” says Jon Bachert, senior director of enterprise applications at Ansira. Because Thunder’s team was simultaneously managing the final leg of the consolidation, it took on the full configuration of the voice assistant—handling architecture, build, and knowledge integration, while Ansira’s IT staff focused on the migration.

The solution runs on Amazon Connect, a cloud contact center service that offers agentic AI solutions for virtually every business function. Amazon Connect draws on a vector store built using Amazon OpenSearch Service, an AWS-managed service that lets users run and scale OpenSearch clusters without managing, monitoring, and maintaining infrastructure. The voice assistant ingests articles from Ansira’s knowledge base—which contains approximately 200 articles covering navigation, order placement, account setup, and campaign guidelines—through the vector store. It then retrieves relevant content in response to caller questions, and delivers answers through a natural-sounding voice built on Amazon Nova Sonic, a speech-to-speech model for conversational AI that delivers frontier intelligence with industry-leading price performance. The solution also uses AWS Lambda, a service that handles all infrastructure management, supporting faster development, improved performance, enhanced security, and cost efficiency. AWS Lambda functions handle the conversational logic, routing callers to a live agent when Sara cannot answer or when the caller requests one. After hours, unresolved interactions create a voicemail-style ticket in Salesforce for follow-up the next business day.

Sara launched in production on the company’s general customer service line, which was chosen because it was not brand specific. That decision compressed the timeline significantly. The build covered more than 50 brands under a single support umbrella, giving Ansira immediate breadth of coverage bypassing more time-consuming, brand-specific rollout. Ansira is among the first companies to deploy Amazon Nova Sonic in a production contact center environment.

Since going live, Sara handles inbound calls across Ansira’s general support line around the clock, including holidays when live agents are unavailable. Initial data shows a call deflection rate of 12 percent, with hold times decreasing by 54 percent. “It has exceeded my expectations,” says Klupe. “Using Amazon Nova Sonic helps our support team to focus on the complex, high-touch partner needs where assisted support adds the most value. As we expand use cases and brands, we expect case deflection to continue increasing.”

The always-on availability addresses a gap that no amount of agent hiring could close affordably. Retail and automotive partners—industries that stay open on the federal holidays that Ansira observes—now have access to support at any hour. With Sara operating on a modular knowledge architecture, expanding to additional brands requires adding segmented knowledge articles rather than rebuilding the assistant. Replication time dropped from a full day of configuration effort to approximately 3 hours per brand group—a 30 percent time savings, reflecting a measurable shift in how Ansira delivers support at scale. “We’ve gone from a significant amount of effort to significantly less, making it possible for Ansira to scale additional brands at a lower operational cost,” says Bachert.