Payter's next step with Agentic AI: from process orchestration to intelligent customer interaction

Rapid growth, smart choices
Payter is a Dutch provider of contactless payment solutions for applications such as unattended vending machines, charging stations, parking meters and donation terminals. Demand for these self-service solutions is growing rapidly, and Payter is growing alongside it.

This growth brings a clear challenge. Customers expect fast service anytime and anywhere, while the organization itself cannot scale its capacity indefinitely. This requires smart decisions in how processes are designed and automated.

Helpdesk under pressure
With the automation of the onboarding process, Payter already took a major step toward scalability. But the work did not stop there. In daily operations, customer inquiries kept coming in. For example, questions about delivery times, how systems work, and what to do in case of technical errors or disruptions.

As Payter grew, pressure on the helpdesk increased rapidly. Employees had to deal with a growing stream of questions, many of which were repetitive. There was clear interest in AI, but not without conditions. The company did not want data to end up in external LLMs and wanted to avoid AI becoming an unpredictable factor in critical processes.

AI enhances, Camunda controls
Together with Incentro, Payter developed an approach in which AI is not used in isolation, but as part of a broader system. The foundation for this was already in place. In an earlier project, Camunda had been introduced as an orchestration layer, centrally managing processes and RPA bots. That same layer is now used to deploy AI agents.

The project started with a pilot that can later be scaled. This pilot took place within the logistics helpdesk, the smallest of Payter’s eleven helpdesks. This environment provided enough variation to learn from, while keeping risks manageable.

Incoming questions, such as delivery inquiries or requests to reschedule deliveries, are analyzed by an AI agent. It extracts the intent from the request, combines it with available documentation and initiates a next step, such as adjusting a delivery date. Where necessary, a human remains involved to review or refine the outcome.

A crucial and distinguishing aspect compared to many other AI projects is that control does not lie with the AI, but with the process. Camunda determines which steps are taken, when an agent is used and when human intervention is required. This keeps the entire system transparent and controllable. Payter can track exactly what happens at every step of the process.

Building on a strong foundation
The collaboration between Payter, Incentro and Camunda builds on an existing foundation. While the first project focused on setting up a central orchestration layer for end-to-end processes, the focus now is on intelligently enriching those processes with AI agents where flexibility and interpretation are required.

Incentro helped translate Payter’s AI ambitions into concrete, workable solutions. By starting small and working iteratively, there was room to experiment while keeping risks under control. Camunda provided the platform that allows AI agents to be integrated as a full-fledged part of processes, including monitoring and control.

Happy customers, happy employees
The first results show that agentic AI contributes to a more efficient and consistent organization. Customer inquiries are handled faster and communication quality is more consistent, as responses are based on centralized and continuously improving documentation.

For employees, this means a shift in their role. Instead of repeatedly answering similar questions, the focus moves toward solving more challenging issues and improving the underlying knowledge base.

Equally important is that Payter maintains full control over the use of AI. By combining AI with process orchestration, visibility is preserved, costs and performance can be actively monitored and adjustments can be made where necessary.

The approach is also highly scalable. What starts within one helpdesk can gradually be extended to other parts of the organization.