- AWS and Nokia are putting AI agents to work on network slicing
- The system moves slicing decisions to the radio layer, using real-time data
- AWS also talked up customer work on network automation
AWS and Nokia are letting AI take the wheel, claiming an “industry-first” iteration of the vendors network slicing technology that is run by Amazon Bedrock-based agents. Operators Orange and du are trialing the new platform.
Amir Rao, AWS’ Director for Global Telco GTM and Solutions, told Fierce that the platform will allow agents to sense radio parameters in real-time to deliver custom network slices. And because Nokia and AWS have already been working to deploy cloud-based 5G Core capabilities, the agents will be able create end-to-end slices.
“It’s beyond the traditional way of looking at network slicing from a 5G Core perspective,” Rao said. “The decisions for network slicing are being made at the radio network operations center, which is the LMS or element management system, which is sensing all the radio network parameters.”
This shift to using real-time readings will allow operators to “anticipate customer needs and deliver tailored services that meet the demands of diverse use cases, from mission-critical to immersive entertainment,” Orange Director of Radio Innovation Atoosa Hatefi said in a statement.
Rao noted Orange and du’s deployments are “not production announcements” but instead are trials of capabilities that will be “eased into production.”
The pair plan to demo the new slicing capability at MWC next week. But network slicing isn’t the only thing on AWS’ MWC agenda.
Automation advancements
AWS Head of Telecom Jan Hofmeyr said that there will be several announcements at MWC showcasing how its customers have implemented AI in their networks to help with automation. Specifically, he said these will focus on work done around change management, mean time to mitigate, fault detection, and node and impairment detection.
That jibes with the results of a recent Nvidia survey which found network automation has become the top overall AI use case that operators are pursuing.
A report from Bain & Company published this month found energy efficiency optimization, service assurance and network optimization are the most mature applications on the automation front, but noted network change management has been harder to implement.
Asked about the primary challenges operators are facing when it comes to automation, Rao ticked off a four-point list.
First is getting a “real-time sense of how the network is behaving,” he said. AWS’ solution here has been the use of digital twins.
Second is the number of nodes.
“This is a really complex problem because even if somebody only has, let’s suppose, 20,000 radio sites in the network…the number of nodes that translate, because those are physical connections, they translate into millions,” he explained. That makes data aggregation and correlation a huge task, but generative AI can help there.
The third piece is the lift of going from ideation into production. “If you think through the network automation lifecycle, you are trying to automate the infrastructure layer, the transmission layer and the application later,” he explained. And that last can’t be done without close collaboration with the likes of Nokia, Ericsson and other vendors.
That’s where point four – working with systems integrators – comes into play, he said. “There’s a capability challenge in the industry as things are ramping up and the AI stack is changing very, very quickly,” Rao said. So it has been critical for partners like AWS to work with telco systems integrators like Amdocs and Slalom to help ensure that the time from “ideation to POC to production can be minimized.”
Nokia will be in Hall 3 and AWS will be in Hall 7 at MWC Barcelona next week.
Catch all our MWC 2026 coverage here.