Ericsson has added a generative AI capability to its NetCloud platform for enterprises to manage their wireless 5G wide-area network (WAN) connectivity. The firm has added a large language model (LLM) in the back-end based on its own library of technical documentation, and a chatbot-style virtual expert in the front-end for enterprise IT teams to converse with about networking issues in their global WAN estates. Ericsson said the new product, called AI-based NetCloud Assistant (ANA), is available immediately to 37,000 enterprises managing 2.9 million devices.
It is the “first generative AI virtual expert designed for enterprise wireless WAN networks”, it said. Instead of trawling the open internet, via open LLMs, the new functionality is geared for enterprises’ particular WAN operations. Its LLM pairs AI insights from the enterprise’s own network with a library of Ericsson’s own technical documentation, and issues personalized responses to prompts from IT departments accordingly, to configure setups, solve issues, and otherwise recommend actions. All the components are hosted “within Ericsson’s environment”, the firm said.
Writing in a blog post, Ericsson explains: “One of the challenges with traditional [gen AI] chatbots is… they leverage API calls to third-party gen AI applications such as ChatGPT. With this approach, it is not easy to distinguish what queries are being answered by the vendor’s AI-chatbot and which queries are being answered by externally-hosted LLMs. This means inadvertent data leaks could occur. ANA is based on LLMs that are entirely hosted [in Ericsson to]… keep all queries internal and away from consumer web-based applications.”
As such, its new generative AI functionality exists in a silo, effectively, between Ericsson and its enterprise customers – “rather than just being artificial generalized intelligence (AGI) like the popular consumer applications”, said Ericsson. More than this, its technical responses go further than just text and hyperlinks, as with standard consumer-geared generative AI applications, to also (“soon”) be able to portray complex data as pie charts, bar graphs, and tables “for easy human consumption”.
These capabilities will use data from Ericsson’s NetCloud data lake and broader WLAN library. ANA works with Ericsson’s NetCloud AIOps dashboard, which detects performance-driven anomalies such as latency and jitter, specific to each customer’s specific environment. It brings automated chatbot support to “read, understand, and infer” information about issues in software-defined WAN infrastructure – to “generate new content” and answers. The end-result is to “[transform] hours or even days of work [for IT enterprise managers] into minutes or seconds”, it said.
Ericsson said: “A current AI innovation is the ability to support an agentic architecture that supports independent AI agents working collaboratively on different tasks. These agents work together using advanced reasoning and planning skills to solve complex, multi-step problems, with LLMs acting as their ‘brains’ for decision-making. The benefit is that it allows for virtual experts to execute multiple tasks in parallel. This allows for more complex responses to be generated and even cross checked, allowing for more accuracy with a faster turnaround time.”
Pankaj Malhotra, head of enterprise networking and security in Ericsson’s enterprise wireless solutions division, said: “Ericsson’s NetCloud is differentiated in its ability to simplify the deployment, management and troubleshooting of enterprise cellular networking. By investing heavily in cutting-edge AI technology, we are empowering even the most streamlined IT teams to tackle 5G administration challenges, enhancing network reliability, security, and user experience with unparalleled efficiency.”
