Ericsson’s CTO sees agents completely transforming enterprise software

Ericsson, the Swedish telecommunications company, has been around for almost 150 years, but that doesn’t mean it shies away from modernization. As a networking company, it uses AI to improve internal operations and enhance its wide array of networking products.

And it didn’t just start using it when it became fashionable after the release of ChatGPT in November 2022. It has a long history of using machine learning and AI to improve network performance, says Ericsson’s chief technology officer and chief strategy officer Erik Ekudden.

Ekudden says that the use of AI inside his organization predates his 2018 start date by a decade, showing the company’s extensive experience with AI. “Today we call it embedded AI, but there has been research in all areas of AI to improve the performance of our products and certainly all our solutions for quite some time,” Ekudden told FastForward.

At Ericsson, embedded AI refers to the concept that the AI is built into products, rather than a separate bolt-on application. But you can’t talk about AI anymore without discussing agents and Ekudden sees some big changes coming as agents become more prevalent in the enterprise.

Putting AI to work

Ekudden's dual role as CTO and CSO puts him at the center of the company's tech and business strategies. It requires working with a diverse group of stakeholders across multiple domains. “We work with the ecosystem that includes everything from research being conducted at universities, to our own research, to how we work with partners and build solutions. We ultimately combine capabilities from the network stack with a cloud stack, and of course, AI nowadays,” he said.

He says the AI strategy has been built over years of development. As the technology has evolved, the company has tried to move with it. “With the [rise of] transformer architectures, large language models, and what came with ChatGPT, we were pretty early in adopting similar technologies to manage and operate networks more efficiently and ultimately create more unique telecom models for network management and network operations,” Ekudden said.

Internally, the company is trying to implement AI as quickly as possible across all business units. Like many companies, Ericsson is seeing the biggest initial success with AI coding applications where Ekudden says the products are a bit more mature than those available for other areas. For other units where the solutions might not be quite as far along, they are working with partners, while encouraging broad experimentation, looking for the best ways to put AI to work operationally across the organization.

“Of course, we build our own tools, and we develop things for our own products, but in-house we're using the best tools that are available. And here I think we were helped quite a lot by having sandboxes that allow people to really try out and jump in, and in some cases build their own, but also to learn from each other and create stronger, faster adoption,” he said.

Agents changing everything

As agentic AI takes on an increasing role in enterprise software, Ekudden wants to position Ericsson to take advantage. For now, that involves incorporating agents into the company’s networking infrastructure, but he also sees agents fundamentally changing how we think about software in the future.

“We have an agentic framework for how we build our products and solutions, easing integration and giving value to our customers,” Ekudden said. This enables the different parts of the network to communicate and coordinate with each other.

That involves an agentic approach for how the company builds products and solutions to ease integration with the network itself, but also with other frameworks, as well as the rest of the IT stack or the cloud pieces that are not provided by Ericsson. He acknowledges that it’s still early and there is a healthy dose of experimentation going on, but he sees agents playing a key role in Ericsson’s products and services sooner rather than later. 

But when he looks at the impact of agents on enterprise software buying, Ekudden also sees a reckoning coming for enterprise software vendors as companies solve the agentic puzzle. He says that companies probably don’t want to be in the business of building bespoke systems, but that if you start to build an organized data foundation and you combine that with agents, it could begin to erode some of the value of some major SaaS businesses.

“It’s almost—well, the question has to be asked—how many of these classical source systems do you want to retain, given that you can actually create so much more value by an agentic framework on top of the data that you have,” he said. It’s worth noting that experts appear to be split on this question. Some believe that SaaS products could evolve to serve AI agents as their primary users instead of humans, while still maintaining their business models. 

Regardless of how it plays out, Ekudden has to continue to help his company navigate these seemingly endless technological shifts, and agentic AI promises to be the biggest one yet.

Featured photo by Daniel Roos, Roos Visual Studio AB