Dutch Telecom Builds an AI System for Millions of Customer Calls

KPN, the leading telecommunications provider in the Netherlands, has developed an AI-powered system to tackle the challenge of handling millions of routine customer calls. In collaboration with McKinsey, KPN created an "agentic" AI solution that goes beyond answering simple questions, enabling the system to complete tasks, interact with back-end systems, and delegate more complex issues to human agents.

This ambitious project required significant changes to KPN's workflows and customer service team dynamics. The firm had already implemented AI chatbots for tasks like explaining bills and answering general inquiries. However, to transform voice conversations into a more useful AI experience, KPN partnered with McKinsey and QuantumBlack, its AI division.

The team began by analyzing vast quantities of anonymized call transcripts and chat records to identify common customer needs and pinpoint which processes could be streamlined. Using these insights, they developed use cases such as customer verification, checking order status, managing technician appointments, and troubleshooting internet issues.

To bring the AI to life, KPN constructed a platform that connected AI agents directly to the company's core telecommunications systems. This enabled the agents to perform more than just provide information; they could now manipulate the system and resolve issues in real-time. The AI was specifically designed for voice interactions, with response times under two seconds per turn and the ability for customers to interrupt agents mid-sentence without disrupting the conversation.

Maintaining human oversight was crucial for safety and quality control. KPN implemented guardrails to ensure AI remained within predefined limits, and established tools to monitor agent performance. Human testers and automated systems evaluated interactions for both quality and safety. Daily reviews of up to 100 call transcripts helped refine the AI's responses, with updates released by the end of each day.

Adapting to this new technology required a workforce transformation. KPN created a task force to update procedures and train employees as AI took on more routine tasks. Frontline staff participated in the development and testing process, while customer-service specialists increasingly focused on more nuanced issues that required human empathy and judgment.

The early results of KPN's agentic AI initiative have been promising. The technology has reduced handling times, resolved some issues without needing technician visits, and allowed human agents to concentrate on more complex problems. KPN aims for the AI to handle 10-20% of customer-service calls by 2027, with customer satisfaction on par with calls handled by human agents.

This case study demonstrates how telecom operators can integrate AI into high-volume customer service operations while preserving the essential human touch required for certain interactions.