Speaking on the ‘Agentic AI: Transforming OSS/BSS with autonomous agents’ panel at Ericsson’s OSS/BSS Summit 2025, the group of industry experts stressed that while early results show promise, clear governance, guardrails and human oversight are essential.
According to BT’s senior leader of AI data and automation in mobile, Girish Mahajan, the industry remains in its infancy when deploying the tool, whilst revealing early success in IT service management, where “there’s a shift from assistive to execution of AI.”
He said: “It has reduced the time of ticket price, it has reduced the time of the manual effort, which is the release of manual efforts, and it has also increased efficiency of the service techs.”
However, he cautioned that “the outcome of authentic AI is something unpredictable,” claiming that as the technology evolves “developers cannot even always predict what an authentic AI system’s outcome will be.”
“We need reflection-based architecture, and we need better AI human collaboration platforms,” he stated.
Meanwhile, Vodafone’s digital and OSS engineering director, Simon Norton, said: “We’ve been working in the AI sphere a long time. “We are one of the very few hands-off, where we have an agentic solution in production,” he said.
According to Norton, that production use case focuses on energy cost recovery and has delivered clear operational benefits. “It acts as one of those lighthouse projects that shows that we can do it, we can deliver value,” he said.
According to Ericsson’s head of product domain data and analytics, Hassan Iftikhar, the company’s experiments began “as a simple dinner conversation with a customer” about simplifying product configuration.
“Very complex catalogue engineering, which can take weeks, is reduced to hours,” he explained. “That was a bit of a step too far,” he admitted, after an attempt to give AI agents autonomous decision-making in operations. We actually had to go in and build guardrails to limit that capability to a human oversight capability.”
Google Cloud’s Aeneas Dodd-Noble, who is the technology giant’s field CTO of telco cloud architecture, notes that while AI agents are now central to its own operations, scaling them introduces new problems.
“When agents start querying the data and trying to bring it all together, all of a sudden, the volume of queries can be overwhelming,” he said, adding that legacy systems often struggle to handle such a load.
However, according to Amazon Web Services’ head of operations, Antonello Arpino, the early phase of AI adoption was marked by enthusiasm but little structure.
“At the start, everyone was rushing,” he concluded. “We built hundreds of pieces of code without the right foundation.” AWS has since shifted focus to “guardrails, human control, explainability of AI, and traceability of AI.”
Related Stories
BT chief warns: AI could drive even deeper job cuts beyond 55,000
Airbus and Ericsson deploy private 5G to boost digitalisation
AI now lies, denies, and plots: OpenAI’s o1 model caught attempting self-replication
