Scaling customer service with Generative AI

The original automation solutions were effective, but the client wanted to keep up with technological innovations and increasingly demanding customers.

Their FAQ bot could handle simple questions, but struggled with nuance. Meanwhile, the backend teams were facing a new bottleneck. The intent recognition system made a huge difference, but the routing logic wasn't granular enough to handle complex, unstructured emails without manual intervention.

In other words: the client didn't just need more automation; they needed smarter automation.

We’ve worked with this client for several years, helping them continually improve their customer service. As technology evolved, we guided the move from simple self-service to fully automated orchestration.

We started by upgrading the self-help experience. To upgrade the standard FAQ search, we developed a generative AI assistant that doesn't just match keywords, but understands context. It generates specific, accurate answers based on the knowledge base, handling a much wider range of questions without human help.

To fix the internal routing chaos, we also overhauled the backend system to create the Customer Contact Assistant. This handles incoming requests in three steps:

1. Categorization: The system uses advanced LLMs to analyze incoming emails and tickets with much higher precision.
2. Summarization: It reads long customer emails and turns them into concise summaries for the support agent.
3. Routing: The system uses complex business logic to send the case to the exact right team, bypassing the general inbox entirely.

These are just two examples of the many initiatives we’ve helped support. Our client is always looking for innovative ways to improve their customer experience. We even conducted an AI maturity scan to map their future ambitions. This helps them prioritize which processes to automate next, so they stay ahead of the curve.

By evolving their automation efforts, our client achieved impressive real-world results. Here are a few facts and figures that stand out:

• Every year, the customer contact assistant automatically categorizes over 150,000 incoming requests.
• Around 80% of the complex cases are now instantly sent to the right team. This eliminates the classic “ping-pong” effect of wrong assignments.
• Every month, the Gen AI assistant handles the first-line queries of almost 10,000 customers.

Customer service agents now spend significantly less time reading questions, and more time solving actual problems thanks to pre-summarized tickets.

To avoid unnecessary migration headaches, the client’s upgraded solution largely runs on the same infrastructure. Azure AI provides the backbone, while OpenAI’s Large Language Models (LLMs) ensure high-accuracy classification. The architecture remains fully GDPR-compliant and now features secure integrations with existing backend systems like SAP and BluePrism to create a seamless flow between the customer and the back office.