- Everyone agrees data is a problem and the first thing to tackle in order to get to Level 4 automation
- For now, humans are still in the loop because the AI makes mistakes
- By 2028, AT&T will have “perhaps over half a petabyte of data per day, collected from our own network," said Joseph Dahan of AT&T.
ERICSSON OSS/BSS SUMMIT 2025, LONDON – Telcos are still drowning in data, and the barrier standing between their success and failure in an AI-led future remains their ability to manage and understand it all, according to speakers at Ericsson’s 26th OSS/BSS Summit.
Here in London, AT&T, BT and wholesale Australian operator NBN all discussed managing their data and how they are using AI to take steps toward Level 4 automation, in which the network automatically makes decisions based on predictive analysis of data.
“Data is the problem, and it’s the first thing we need to tackle,” said Phil Stubbs, AI/Data software engineering director at BT Group. “There is more data produced today than anyone could possibly use.”
The data problem is no surprise to industry insiders — to quote Jane Austin’s Pride and Prejudice, “It is a truth universally acknowledged."
Alex Bhak, global lead of the telecommunication practice at Bain & Company, also hammered on the same point, as did Joseph Dahan, director and member of the technical staff at AT&T.
What’s different this time around?
This Fierce Network reporter has been hearing about “managing the firehose of data” since at least the beginning of the millennium. What’s different, though, is AI. If used prudently, it could finally deliver increased revenue and lower capex and opex for telcos.
It's a massive problem — in the exabyte category.
Dahan said he estimates that by 2028, AT&T will have “perhaps over half a petabyte of data per day, collected from our own network.” As that data grows, the silos within a telco also grow.
“I think everybody probably has the same issue, where data solutions have been built over the years and you have fragmented data sources, fragmented databases,” he said. “The strategy focuses on open, interoperable, secure and manageable data formats to avoid silos and ensure data lineage and hygiene.”
Stubbs said that telcos can’t make do by adding more people, instead he thinks the industry can apply generative AI (GenAI).
"If we keep adding more people, we will never catch up." The answer is that operators have to "change the people." He added, "We have to teach machines how we want them to behave."
Interaction with the data
NBN has figured out how to use AI and GenAI to its advantage when managing network data, according to Richard Kendall, enterprise GM of architecture. The wholesale operator looks at the interaction between the data, the processes and the different GenAI tools, platforms and features for automated decision making.
Kendall said the provider had "great success" bringing together disparate data platforms into a single "fabric" used across the company but humans still had to be in the loop to add context back in, which slows down the process.
Later in a one-to-one briefing, Ericsson's Jason Keane, head of BOS portfolio, concurred that humans indeed still need to be in the mix.
"You need machine learning to predict something going to happen; GenAI to answer the question, 'why?''; and agentic AI to tell you how to fix it," he said. "Agentic AI can give you 10 steps on how to fix it, but it makes mistakes. A human has to be involved still in the loop."
So will AI solve the data problem? If used prudently, maybe. Telcos can hope. As Orange Business' Usman Javid recently told Fierce Network Research's Chief Analyst Mitch Wagner, AI is "one of the few lifelines left for the telco industry."