Autonomous AI in Payments: How Alma Scaled Level-1 Support Without Compromising Compliance
As BNPL volumes grow and support peaks intensify, Alma automated Level-1 customer support with autonomous AI agents while maintaining accuracy, compliance, and trust.
6,500
Level-1 tickets handled autonomously per month
-3 pts
Re-iteration rate compared to human-handled tickets
-40%
Reduction in full resolution time
About the Client
Industry
Fintech
Location
Paris, France
Company
Alma is a European fintech company specializing in Buy Now, Pay Later (BNPL) solutions, enabling consumers to split purchases into multiple instalments while allowing merchants to offer flexible payment options without added risk.
Alma partners with more than 23,000 e-commerce and retail merchants across Europe, and over 9 million consumers have used Alma at least once throughout the full lifecycle of their instalment payments, from checkout to final repayment.
With a strong consumer-first approach (no late fees, transparent payment plans, and a 4.8/5 rating on Trustpilot) clarity, accuracy, and responsiveness in customer support are critical to maintaining trust and long-term adoption.
Business Context
As BNPL adoption accelerated, customer expectations around support evolved rapidly. Consumers expect instant, precise, and reliable answers when asking about payment status, instalment schedules, failed payments, or next steps.
At the same time, Alma’s support organization faced a growing volume of inbound requests driven by:
- Replies to outbound payment communications (upcoming instalments, reminders, failed payments),
- Recurrent questions about payment management, including instalment schedules, flexibility around payment plans, refunds, or next steps following a failed payment.
Most of these interactions followed strict, well-defined business processes, yet remained highly repetitive and time-consuming for human agents.
Alma identified autonomous AI as a way to scale Level-1 support, while ensuring:
- Full compliance with internal processes and financial regulations,
- Zero hallucination or off-policy responses,
- Smooth collaboration with human agents for escalation and edge cases.
The Operational Challenges
As Alma continued to grow rapidly, its support team faced a structural challenge linked to increasing contact volumes driven by adoption and scale:
- A large share of inbound tickets could be resolved using existing data and processes, but still required manual handling.
- Support volumes increased alongside payment cycles and user growth, making it inefficient to scale human staffing linearly while maintaining consistent service quality.
- Multilingual requests increased operational complexity while requiring the same level of precision and compliance.
- Human agents spent a disproportionate amount of time on Level-1 tickets, limiting their availability for complex, sensitive, or exception-based cases.
The objective was clear: automate Level-1 support end-to-end, without sacrificing accuracy, compliance, or customer experience.
The AI Solutions
What was deployed
Alma deployed a process-driven autonomous AI agent, powered by Yampa and fully integrated into its customer support stack, including Zendesk, customer data systems, and internal knowledge bases.
The Yampa agent acts as a true Level-1 support layer. It retrieves payment information, applies Alma’s documented business processes, and delivers accurate, compliant responses to standard customer requests, while escalating to human agents whenever required.
The deployment followed a phased and iterative approach. Alma started with a limited set of well-defined contact types, then progressively expanded the scope as performance and confidence increased. Over time, this included additional processes, languages, higher volumes, and more complex multi-touch interactions.
This approach allowed Alma to industrialize automation while maintaining strict control over quality and compliance.
How the agent operates
- Customer emails are received through Alma’s existing support tools.
- The AI identifies the customer, analyzes intent, and maps each request to the appropriate documented process.
- The process is executed using Alma’s customer data and internal knowledge base.
- A compliant, personalized response is generated in the customer’s language.
- When a request is unclear, out of scope, or sensitive, the ticket is escalated to a human agent with full context and clear reasoning.
This setup enables end-to-end automation of standard support use cases, far beyond simple FAQ handling, while preserving seamless collaboration with human teams.
The phased rollout ensured tight operational control, safe scaling, and alignment with Alma’s growth trajectory.
The results
- Human agents refocused on complex, sensitive, and high-value cases, improving overall support quality.
- Consistent, 24/7 compliant responses delivered across languages, aligned with Alma’s internal processes.
- Greater operational predictability, allowing Alma to absorb payment-cycle peaks without over-staffing.
6,500
Level-1 tickets handled autonomously per month
-3 pts
Re-iteration rate compared to human-handled tickets
-40%
Reduction in full resolution time
"Yampa now automates a significant share of our Level-1 support and frees our teams to focus on complex cases. The collaboration has been extremely smooth, iterative, and grounded in a consumer-centric use of AI, without compromising accuracy"
Want to automate level-1 payment support like Alma?
