• Five benefits are driving AI transformation at the operator
• Agentic AI has the biggest value, and has led to purpose-built governance frameworks
• Mutual learning is essential to exploit AI-native opportunities in telcos

Düsseldorf, Germany – AI-Native Telco Forum – Deutsche Telekom (DT) has outlined the rapid progress it’s making in transforming its network operations using multiple AI capabilities.

Speaking at the AI-Native Telco Forum – which is organised by TelecomTV and co-hosted by the operator – senior vice president for network strategy & data and AI in networks at DT, Ahmed Hafez, highlighted the scale of the opportunity for telcos and the need for mutual learning on this fast-evolving subject.

“There’s a chance to be overwhelmed,” said Hafez. “We have to be humble around AI – no-one can manage alone.”

Hafez described how developing autonomous networks with AI at the centre would provide five main benefits: Improved customer experience, faster time to market, lower total cost of ownership, improved network resilience and the opportunity to develop new capabilities.

“We need to care about resilience. We need to be robust against security hazards and outages,” said Hafez. “To be resilient without AI will cost a significant amount of money – AI has huge potential.”

Deutsche Telekom already has 150 use cases for predictive AI but the scale of generative AI use cases is going even faster, said Hafez. The operator is now making huge strides with agentic AI, which has “the biggest value” to the operator.

Hafez outlined how the operator has developed a strong AI governance capability that mitigates the potential risk to the network of increasing autonomy, explaining that Deutsche Telecom has a human team to verify and review the actions of agents but that for most actions, the operator was happy to let AI agents make the decisions.

Risk can be mitigated by giving the agent a finite list of actions to choose from, all of which are safe. “We have humans managing agentic AI and monitoring the results,” he said. “We can stop the AI at any point.”

And there is automated transparency too, with the agent logging what happened as well as the result. 

Hafez stressed the need for a common language, given the fast pace of change in AI capabilities, and noted that this has been implemented within Deutsche Telekom and its suppliers. However, its definitions will evolve over time as new opportunities to develop AI arise, he added.

– Richard Thurston, Contributing Editor, TelecomTV