- AI-powered automation is not the same thing as an autonomous network, Ericsson’s Mats Karlsson told Fierce Network
- Level 5 autonomy is likely to emerge selectively, while humans will continue to supervise many closed-loop processes, he said
- Investing in higher-level autonomy needs to produce a business return
ERICSSON OSS/BSS SUMMIT 2026, ZURICH, SWITZERLAND – Telecom operators have been automating their networks for years, and the arrival of GenAI doesn’t suddenly make those networks autonomous, Mats Karlsson, VP and head of solution area operations and business support systems at Ericsson, told Fierce Network.
“Everything that relates to AI, [some operators] believe it's autonomy,” Karlsson said. “AI can be used both for doing... hyper automation, but it's not the same as autonomy.”
The difference, he said, is intent. Traditional automation can accelerate an existing process. Autonomy starts with specifying the desired state of a network or service and allowing the system to determine how to reach it.
Where do humans fit in network automation?
TM Forum created an industry benchmark for network automation six years ago when it created five levels of autonomous networks:
- Level Zero is all manual network operations
- Level 1 is assisted operations
- Level 2 is partial autonomy
- Level 3 is conditional autonomy (humans still providing oversight)
- Level 4 is high-level autonomy, or intent-driven, predictive and closed-loop AI/ML management. It can self-heal, self-optimize and self-configure across complex, multi-domain environments with minimal or no human intervention for routine operations. This level is the primary target for most operators
- Level 5 is full autonomy.
Level 5 autonomy barriers
Karlsson expects fully closed-loop Level 5 autonomy to emerge selectively, for specific use cases where operators are confident enough to remove human involvement. “I think closed loop will be very much a stepwise journey,” he said.
Operators also face practical barriers. Karlsson pointed to the need for real-time understanding of network state and topology, as well as organizational changes as decisions traditionally handled by engineering teams, operations groups and change-control boards move into closed loops. Existing OSS/BSS architectures also need modernization to support Level 4 and Level 5 autonomy, he said.
Does higher autonomy mean better business?
Higher autonomy isn't automatically better business.
Karlsson said industry autonomy levels provide a useful common language but cautioned against turning them into a numbers game. "It's not an automatic translation to the business impact,” he said.
A Level 4 investment for a task performed twice a year may never pay for itself, while automating something executed 1,000 times a day presents a very different equation.
Nearer-term opportunities are more concrete. For example, Karlsson said fault management is an early target, with service assurance, customer experience, quality and performance optimization emerging as the “next wave.”
Operators can also achieve substantial benefits at Levels 2 and 3, he said, without racing immediately toward full autonomy.
In other words, the meaningful benchmark may ultimately be less about achieving Level 5 and more about whether autonomy solves a problem worth paying to solve.
Read more about network automation and Ericsson on Fierce Network
Verizon pushes autonomous networks toward the customer