- Telco leaders say automation and autonomy are often conflated
- T-Mobile is pursuing autonomy around customer outcomes rather than chasing full domain-level autonomy
- AWS warns operators still need stronger software deployment practices and AI governance to get to full autonomy
DTW IGNITE 2026, COPENHAGEN, DENMARK – Autonomous networks are automated with AI but not all automation equals autonomy. It might sound like wordplay, but for vendor and operator executives it’s a worthwhile distinction to make at a conference where both terms – automation and autonomy – are being thrown around like rice at a wedding.
Anders Vestergren, head of Solution Area Network Management, BCSS at Ericsson, told Fierce the difference between automation and autonomy is like the difference between a car with an automatic gear box and a fully self-driving car. One simply makes the process of driving easier by automating the action of shifting gears. The other actually understands the intent – for instance, I want to go to grandma’s house – and takes all of the necessary steps to get you there.
Achieving autonomy – either of a single telco domain like RAN operations or across all domains – isn’t easy by any stretch of the imagination. Even on a process level, the difficulty level depends on the use case, Vestergren said.
“There are Level 4s now. For some use cases, Level 4 is reasonably easy. For some use cases, Level 4 is very difficult,” Vestergren said. “In energy management, it’s occasionally easy to be Level 4 because it’s a single domain and the autonomous features which are enabling it are mature and well-known.”
Domain vs process autonomy
But not everyone agrees that full domain level autonomy should be the goal.
T-Mobile Chief Network Officer Ankur Kapoor told Fierce that he’s focused on using autonomy as a means to achieving specific customer outcomes. And that doesn’t necessarily require achieving vertical domain-level autonomy, but instead creating autonomous processes horizontally to facilitate a particular customer experience.
“If I can do a lot of autonomous operations in the core, that’s a win, right? That helps a lot with network operations, but it doesn’t help much with overall end-to-end service experience.”
Telcos have historically focused on coverage, speed and latency. “I think now is the time and era in this autonomous journey from where we are moving from those to actual outcomes,” he said. That is, a customer doesn’t care how they get a service, they just care that it works. So, T-Mobile is focusing on using automation and autonomous processes to make that happen.
Bumps in the autonomous road
Regardless of whether operators are pursuing process or full domain autonomy, there’s still a long road ahead.
“I think there’s a lot of work to be done to get the network to a point where just the basics around software deployments can be done,” AWS VP and telco chief Jan Hofmeyr told Fierce. “I tell everybody, you need to get to a point where you can deploy code every single month – at a minimum – and have a plan that you can, if there was a zero day vulnerability tomorrow, that you can deploy same day.”
Those kinds of baseline capabilities will become increasingly important in the AI era, especially as the technology changes the threat landscape. Hofmeyr said companies who can’t respond in this manner risk being “consumed” by threats and having to shift all their attention away from innovation in favor of defense.
Hofmeyr said if you ask operators today, they’ll say they’re in a good place as far as software deployment goes. But he noted that code deployments are not yet touchless for many operators and that the human-in-the-loop approach isn’t likely to change anytime soon.
To move towards more autonomous operations, Hofmeyr said there will need to be more declarative guidelines around what AI agents can and cannot do.
“It sounds like governance, but it’s more than governance,” he said. “Will you let an agent delete something from a database? Not anytime soon. That’s a destructive event. Even in today’s world, when we do that as humans, there are so many checks and balances that happen before we could delete something from a database. So, you need all of those same guardrails when AI is allowed to do that. But we are not close to that.”
Incidentally, humans themselves are also proving to be a hurdle on the road to autonomous operations.
“The largest stumbling block is the people, the ways of working and the process adaptation,” he said. “Of course, there are barriers in technology, too, but in 70% of cases, technology is used as a scapegoat.”