Ericsson launched an agentic rApp service on Amazon Web Services (AWS), expanding its relationship with the hyperscaler alongside its open radio access network (open RAN) footprint.

Exclusive to AWS Marketplace, Ericsson’s Agentic rApp as a Service (rApp aaS) is promised to enable efficient, scalable, and flexible deployment of rApps – microservices-based applications operating in non-real time. Hosted on AWS, the service connects to Ericsson's non-real time RAN Intelligent Controller (non-RT RIC ), within the service management and orchestration (SMO) framework over the Open-RAN (O-RAN) Alliance R1 interface – the defined standard for connecting third-party applications to a communications service provider’s (CSP) network.

As expected in the post-ChatGPT area, the solution also integrates a natural language interface, which invites CSP teams to interact directly with the network by converting natural language inputs into system-level instructions understood by the overall agentic AI framework. As such, Ericsson framed the service as augmenting the network optimization process by leveraging AI for advanced reasoning and coordinated workflow management.

Field testing of the service is underway with CSPs like Vivo Brazil, whose director of mobile network core and deployment, Elmo Matos, said: “Collaborating with Ericsson on this initiative allows us to test different applications of artificial intelligence aimed at enhancing network experience. With AWS support, we will explore the rApp aaS model and expand AI-driven functionalities, focusing on improving network capacity and automation on a national scale.”

No further details were given of the Vivo partnership, for example gains seen with the use of AI deployments, but Ericsson claimed in its announcement to currently handle more than 100 million AI inference actions daily across 11 million cells serving more than two billion subscribers.

Regarding Ericsson's AI ambitions, Daryl Schoolar, analyst and director at Recon Analytics, said the firm needs to clarify what AI offers in the RAN space.

"AI is such a heavily used term these days in technology, Ericsson needs to continue to educate the industry on the three different approaches to AI and RAN that make up AI RAN – AI applications for RAN, AI built directly into the RAN software stack to improve performance, and AI on RAN where AI applications are co-located and run on RAN infrastructure.

"AI applications for RAN is well understood, but Ericsson needs to build a story about the benefits and deployment scenarios for when AI is built into the software stack and when AI applications are co-located and run on RAN infrastructure," Schoolar continued. "It is important this year that Ericsson shows real-world trials of these AI RAN approaches on live commercial networks to validate their approach and to build mobile network operator interest."

Partners on the operator and hyperscale side

Away from AI and on the more bread-and-butter side of Ericsson’s rApps, past operator partnerships have included AT&T, who last summer deployed a third-party rApp on its live production network in a brownfield environment, touted as the first example of a non–real time RIC commercial deployment.

The deployment allowed the rApp to communicate with the SMO layer using open, standardized APIs, highlighting the potential of open standards in disrupting closed, single-vendor self-organizing networks (SON).

Joe Madden, lead analyst at Mobile Experts, told SDxCentral the AT&T effort underlined Ericsson’s leading status in open RAN.

“It may surprise people to hear that Ericsson is the leading vendor in the commercialization of open RAN software," Madden said. "Ericsson resists some features of open RAN that don’t support their vision of high-performing networks. But they have established an ‘app store’ of open RAN optimization applications from more than 20 different developers.”

“It also controls more than 40% of the global mobile network outside of China, so any applications developed for Ericsson networks have a large market potential," Madden added.

Ericsson's AWS announcement also saw the vendor tout its established history with the hyperscaler. The Swedish giant offers its operations support systems (OSS) and business support systems (BSS) on AWS, while its rApp launch builds on previous agentic-driven autonomous network ventures between Ericsson’s Cognitive Network Solutions and AWS. Other joint AI efforts also included a generative AI lab which has worked on agentic-powered OSS/BSS solutions for the likes of Indian operator Grameenphone.

In an interview last year, Jason Keane, Ericsson’s head of OSS/BSS portfolio, told SDxCentral that the firm’s advantage comes from its exclusive openness to hyperscaler partnerships.

“The hyper-cloud providers have the platform and technology, and we'd rather partner with them," Keane said. "We sell our expertise and we just package into software and everything else, so that’s where I think where we're different. Ericsson can bring your full end-to-end network where out rivals can’t.”

Away from Amazon, Ericsson has built a carrier-grade 5G core as-a-service with Google Cloud, in addition to the validation of 5G core network functions on Microsoft’s Azure Operator Nexus platform.