01 — CHALLENGE

Processing 8,000 Tickets Monthly and Spending 20% of the Time on Informational Requests

Although high ticket volumes are common in B2C businesses, Bancontact Payconiq Company's support team reached a critical breaking point in their customer service delivery. The situation became particularly challenging during seasonal peaks throughout the year, as the team size remained constant despite fluctuating demand. They managed approximately 400 tickets daily, leading to extended response times. Nearly a quarter of incoming enquiries lacked proper categorisation, whilst 20% were purely informational requests that didn't require agent intervention. "We were managing a large number of requests," recalls Martijn Toch, Consumer Care Agent. "The combination of backlog and repetitive questions made it clear that many enquiries didn't require direct agent involvement."

‍

‍

02 — SOLUTION

Automated Ticket Sorting and Response with Penny, an AI-Powered Assistant

As BPC's trusted partner, we suggested developing and implementing an AI assistant to automate initial response generation through intelligent ticket categorisation, personalised template-based responses. We analysed the most impactful support cases and identified three key types of requests to address:

- requests where users only need to be informed

- requests where it's important to react urgently

- requests which appear often

Quality control is maintained via a human-in-the-loop verification system with robust quality assurance checkpoints and customisable response templates. "The solution tackled our two main challenges: uncertain email categorisation and handling informational requests that didn't require agent intervention," explains Maya Vanard, Project Manager. Penny was designed to excel in high-impact scenarios, offering specialised handling for urgent cases while processing common app usage and payment enquiries and automating information-only requests.

‍

‍

03 — APPROACH

Building a Robust, Scalable Solution Leveraging Modern Cloud Technologies

Penny's integration involved sophisticated architecture seamlessly connected to Bancontact Payconiq Company's existing systems. The workflow is elegantly simple: "Questions enter Jira, triggering Penny's analysis. The system then generates a proposed response directly within the ticket," shares Martijn.

The team created templates for common scenarios and established a regular system refinement process to ensure continuous improvement. The workflow design established ticket capture through email and app interfaces, utilising AI-powered categorisation and automated response generation with personalisation, all subject to human review.

Penny’s technical architecture leverages Azure Cloud Infrastructure, with backend services running on Azure App Services and OpenAI models handling natural language processing. Azure Table Storage enables comprehensive performance monitoring and analytics. The system seamlessly integrates with the BPC's existing infrastructure through direct connections to their Jira environment, Confluence knowledge base and custom API endpoints for real-time processing.

The collaboration between In the Pocket and Bancontact Payconiq Company goes beyond typical client-vendor relationships. "We worked with complete transparency," Maya notes. "Even in the change management part of the adoption, they provided practical solutions and support." Rather than rushing to add features, the team focused on perfecting core functionality. "Excellence before expansion became our mantra," she adds.

‍

‍

04 — OUTCOME

Faster Response Times + Efficient Agents = Happier Clients

Penny’s key features included intelligent categorisation with automatic ticket classification and pattern recognition for common requests. The response system generated contextually aware, personalised responses with appropriate empathy and multi-language support. Comprehensive performance monitoring tracked usage, response acceptance rates and agent satisfaction alongside system analytics.

When implementing AI tools, it's crucial to consider both the learning curve and the human element, as this requires continuous work in terms of adoption. During Phase 1 of our implementation, only 17% of agents reported that the tool provided added value—a clear signal that we needed to focus on improving usability and user experience. We modified our approach, focusing on usability to seamlessly integrate it in their current day-to-day workflow to increase adoption. Our analysis of the latest phase showed significant improvement: 88% of agents now confirm that Penny delivers valuable assistance.

The impact has been remarkable. Combined with other optimisation tools, the system reduced the backlog from 8,000 to near zero, while response times improved to 1-2 working days. The faster responses improved significantly with built-in empathy, and agent satisfaction reached new heights. Agent feedback has been overwhelmingly positive. "Penny is now irreplaceable," shares Maya. "When asked how much they'd miss the system, agents rated it 8 out of 10."

The platform's success led to strong adoption rates and continuous expansion of use cases and secured additional investment for AI development. The project continues to evolve, with plans to enable Penny to respond directly to customers. "We're laying the groundwork for direct AI-customer interaction," says Maya. "But for now, we're focused on continuous improvement of our current system."

The collaboration was successful not only because of the innovative solution but also because of transparency and empathy. "During our weekly meetings, there was a genuine interest and I felt heard. The In The Pocket's team didn't just listen—they understood and delivered solutions that transformed our daily operations," concludes Martijn.