Ninety-four per cent of broadband service providers plan to deploy software-defined networking as they move toward autonomous operations, but 83 per cent say inconsistent data from existing systems blocks them from running AI effectively, a Broadband Forum survey shows.

The October 8 report polled operators on their path to autonomous networks. Ninety-two per cent described AI and machine-learning integration as critical or important to their strategy.

What is blocking autonomous broadband?

Poor device compatibility and fragmented data top the list. Seventy-eight per cent of respondents named poor interoperability between vendors as a major barrier. Ninety per cent identified integration with legacy infrastructure as a key challenge.

That legacy footprint stretches across DSL, cable/HFC and multiple generations of fiber equipment already in the field.

Predictive maintenance dominated near-term AI plans. Nine in ten providers said they use or intend to use AI/ML for predictive maintenance. Their other priorities:

• Anomaly detection (72%)
• Automated fault resolution (69%)
• Customer experience optimisation (87%)
• Service assurance (85%)

How do standards fit in?

Vendors and operators cite standards as the connective layer across multi-vendor networks. Broadband Forum chief executive Craig Thomas separated execution from intelligence in the autonomous model.

"A software-driven foundation gives networks the ability to execute an instruction on its own, while AI provides the intelligence to let it make decisions and adapt to changing conditions automatically," Thomas said.

Operators must reconcile hardware, software, cloud platforms and management systems from different vendors before automation can scale, he added.

What role should vendors play?

Radisys senior product manager Robert Soukup argued that standards bodies must engage beyond writing specifications.

"Taking part in reference implementations, interoperability testing, and proof-of-concept programs enables service providers and vendors to validate common approaches against real network lifecycles and feedback practical deployment experience back into the standard," Soukup said.

What happens next?

Operators' near-term priorities line up with the report's most mature AI use cases. Predictive maintenance and anomaly detection rely on narrower, well-defined data inputs, which let providers deploy them while broader integration work continues.

The timeline now hinges on how quickly vendors move on interoperability testing. With 78 per cent citing vendor fragmentation as a major barrier, the pace of plugfests and reference implementations will shape how quickly operators extend AI beyond isolated maintenance workflows into full service assurance and customer experience use cases over the next 12 to 24 months.