Atlanticus scaled to over 159,000 monthly interactions and a 60% automation rate without ever losing sight of the customer on the other end.
Atlanticus uses technology and analytics to help bank partners deliver financial solutions for everyday consumers. That mission makes every service interaction count: customers need answers that are accurate, easy to understand and specific to the account task they’re trying to complete.
With boost.ai, Atlanticus built its AI Agent, Sandy, to do more than field frequently asked questions. Deployed across seven of Atlanticus’ brands and connected to more than 80 integrations, Sandy supports a wide range of service journeys – from PIN requests and card replacements to auto-pay enrollment, late-fee waivers, and payments – while connecting customers to a live agent when that’s the better path.
In August 2026, Sandy handled over 159,000 conversations and serviced nearly 60% of them without transferring to a human. A result that strongly reflects a way of working that treats improvement as an ongoing mission, rather than a one-time goal.
Atlanticus’ results at a glance…
-
60% of August 2026 conversations automated by Sandy without transfer – up 5% from January 2025
-
159,000+ chat conversations in August 2026; 75,436 unique customers served without transferring to live chat
-
$2.96 million in payments collected YTD across more than 20,000 transactions
-
86% training success score, above Atlanticus’ 85% target
A foundation built on continuous training
Atlanticus’ automation performance is founded on boost.ai’s Hybrid NLU approach – which combines a custom-trained intent prediction model with LLM technology to give teams a high degree of control over how the AI Agent understands customer messages – combined with an operating model that treats improving the AI as continuous work. It's an approach built for financial services: boost.ai's platform is purpose-built for regulated industries, where accuracy and compliance aren't optional. The team doesn’t treat feedback as a score to report and move past; they review why interactions succeeded or went unresolved, and use those findings to drive tuning and customer journey design.
“Our results combine the boost.ai platform with disciplined oversight and a clear understanding of what our customers need. The Atlanticus team continuously applies insights from real interactions to strengthen performance and eliminate friction. That commitment has allowed us to deliver automation at scale while elevating the customer experience across our brands.” - Dari Damazo, SVP of Support Services, Atlanticus
Negative feedback, specifically, gets a closer look. Atlanticus separates bot-related dissatisfaction (i.e., missing intents/topics, insufficient training, etc.) from policy-related dissatisfaction, where Sandy delivered the correct answer but the customer didn’t like the outcome. The distinction keeps the team focused on what’s actually within reach and ensures that customer frustration still informs the broader experience even when Sandy did its job correctly.
Turning a friction point into a breakthrough
In March, 2026, Atlanticus saw immediate escalations rising – customers opening a chat session and requesting a live agent before Sandy had a chance to help. The team responded by redesigning Sandy’s welcome experience using two boost.ai platform capabilities: context topic, which allows Sandy to track the subject of a conversation across multiple messages, and entity extraction, which pulls specific information – like the reason for a customer’s request – directly from what they type.
Customers who asked for an agent were immediately invited to share more context and give Sandy an opportunity to help. If they repeated the request, they were then transferred to a human agent, without further friction. The change was careful and monitored: Atlanticus tracked both operational outcomes and customer feedback to make sure the adjustment wasn’t shifting frustration instead of reducing it.
Immediate escalations fell by 46% from March to April and stayed consistent through August, with Atlanticus reporting no increase in negative customer feedback tied to the change. The pattern reflects the team’s approach: identify friction, protect customer choice, and monitor both operational and experience outcomes.
“The opportunity ahead is to make every customer interaction more intelligent, connected, and effortless. Boost.ai understands both our business objectives and the standards required in financial services, making them an important partner as we continue to evolve Sandy. Together, we are creating a foundation for responsible innovation, sustainable growth, and an even stronger customer experience.” - Dari Damazo, SVP of Support Services, Atlanticus
Self-service that completes and informs
Sandy’s value is clearest when a customer finishes a task quickly and confidently. Customer feedback about the AI Agent describes it as “much easier and faster than calling,” “seamless,” and “quick and easy” – across interactions like card support, PIN guidance, autopay enrolment, and late-fee waiver assistance, which is a task that previously required live agent assistance and is now fully automated through Sandy.
Payments show the same shift from information to action. Built-in guardrails in the boost.ai platform ensure every interaction stays within regulatory boundaries, making it possible to handle sensitive transactions like payments safely and at scale. Sandy processes Automated Clearing House (ACH) and debit payments end-to-end within the conversation, resolving a need in the moment rather than directing customers to another channel. Through August 2026, the AI Agent had reported more than $2.96 million collected across more than 20,000 payments. For customers and service teams alike, that’s a meaningful reduction in effort on both sides.
What’s next for Sandy
Looking ahead, Atlanticus is exploring generative AI capabilities – building on the self-service foundation already in place and expanding the range of interactions Sandy can handle. With the integrations, the AI training discipline, and the operating model already established, it’s an evolution the team is well-positioned to make.
“What excites me most about what comes next is that we are building from a strong foundation. Sandy already completes meaningful transactions at scale, and our next step is to explore generative AI in a way that is controlled, compliant and genuinely useful to customers. Boost.ai has been a true partner in that journey, helping us balance innovation with the accuracy and trust our customers and banking partners expect.” - Nicole Avery, Support Services Specialist, Atlanticus
The combination of the boost.ai platform’s strong foundations, continuous customer feedback, and controlled, high-value journeys are what turns an AI Agent into a trusted service channel. Atlanticus has built that combination – and shown what happens when a team is genuinely committed to making it work.
