KPN tackles millions of calls with AI

The Dutch telecom giant handles roughly 5 million customer calls yearly. It partnered with McKinsey to build an AI system that answers phones, completes tasks, and routes complex issues to humans.

AI learns from past conversations

KPN analyzed millions of anonymized call transcripts to identify what customers actually need. This shaped which tasks the AI could handle: verification, order status, appointment booking, and troubleshooting.

The system talks like a human

KPN built a voice AI with response times under two seconds and the ability to understand interruptions mid-sentence. It connects directly to KPN's core systems so it can actually complete tasks, not just answer questions.

Guardrails keep the AI honest

KPN built safeguards to keep AI agents within boundaries and monitors their performance daily. Human testers and automated checks evaluate interactions for quality, and teams review up to 100 calls daily to refine the system.

Employees adapted faster than expected

KPN created a task force to retrain staff as AI took over routine work. Frontline employees helped develop and test the system, then shifted to handling complex issues. Employee adoption achieved over 86% success.

Early wins and expansion plans

The AI reduced handling times and resolved some issues without technician visits. Customer satisfaction matched human-handled calls. KPN aims for AI to handle 10 to 20 percent of calls by 2027.

Redesign matters as much as technology

KPN's success came from rethinking workflows, training staff, and building oversight alongside the AI. The lesson: deploying agentic AI requires operational redesign, not just new software.

“KPN's experience shows that putting agentic AI into customer service involves as much operational redesign as technology deployment.”