This week The Prompt Economy weekly dials up the telecom sector. It’s here that developers and engineers are exploring new ways to incorporate agentic artificial intelligence to fix problems and create a better customer experience.
For example: Nokia says AI agents can help telecom engineers find and fix network problems faster. In an Oct. 27 article, Nokia’s Martin Holecek describes a system that lets engineers request help using everyday written or spoken language. The system brings together information and tools that engineers would otherwise check separately. Nokia says it’s already using this approach in its network design and improvement services. The goal is to reduce time spent gathering information and give engineers more time to resolve service problems.
Dropped calls offer one example of how the approach could work. An engineer could ask an AI agent to track call failures and investigate when they exceed an acceptable level. The agent could identify likely causes, recommend a fix and carry it out with human approval. It could then check the results and summarize what changed.
Nokia argues that this approach could reduce repetitive work and help teams apply expert knowledge more consistently. Engineers would retain control over consequential actions. The article doesn’t provide measured results, so the expected gains in speed and service quality remain unquantified.
The Developer View From Nvidia
Engineers and developers, as the Nokia report shows, are the targets at this point for agentic deployment. Telecom companies need shared data, tools and clear operating rules to give AI agents more responsibility, Nvidia argues in a recent technical blog post. Many operators still use automation mainly to carry out predefined tasks. Nvidia describes a broader role for agents that investigate unfamiliar problems, compare possible fixes and track results over time. Some agents would handle individual requests. Others would monitor networks continuously or research better ways to operate them. They’d share information and work within the same security controls. Operators would review higher-risk changes before agents put them into effect.
Nvidia illustrates the approach with a simulated network problem. When traffic slows or a connection fails, one agent investigates the cause and ranks possible fixes. Another applies a plan, checks whether service improves and tries an alternative if needed.
The company also describes early research into using agents to develop better wireless network algorithms. Nvidia reports that one experiment improved the efficient use of wireless spectrum by more than 3% over an industry-standard method. These examples show potential uses in both maintenance and research. They don’t establish results across live telecom networks. For operators, Nvidia’s recommendation is to start with a valuable task and build on shared systems as more uses emerge.
Using Agentic AI in the Real World
Dutch telecom provider KPN is using AI agents to handle routine customer service tasks while keeping human support available, according to a McKinsey case study. KPN worked with McKinsey and its AI business, QuantumBlack, to develop voice agents for tasks such as verifying customers, checking orders, managing technician appointments and troubleshooting internet connections. The company handles roughly 5 million customer calls a year. It aims to have agentic AI handle 10% to 20% of service calls by 2027. The case study reports a customer satisfaction score of 83, comparable to calls handled by people.
KPN’s experience shows how much work goes into making AI service feel natural and dependable. The team designed conversations around response times below two seconds and let customers interrupt agents midsentence. Staff reviewed transcripts from up to 100 customer calls each day and used the findings to improve the agents. Frontline employees also helped develop and test the service. McKinsey reports shorter handling times and more efficient problem resolution, though it doesn’t quantify those gains. KPN is also training its own teams to develop additional uses. That gives the company a foundation for expanding AI service while employees focus on more complex customer needs.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
