Juniper Networks takes on complex legacy network access control (NAC), introducing its own cloud-based, Mist AI-driven solution. It also enhanced its Marvis virtual network assistant (VNA) with the ChatGPT integration.

Traditional NAC services built using on-premises overlay hardware are often complex and brittle to deploy, operate and authenticate the devices, while lacking agility and ability to scale, noted Christian Gilby, senior director of product marketing at Juniper Networks.

Juniper’s Mist Access Assurance service is built on the technologies from its WiteSand acquisition. The networking vendor bought the cloud-native, zero-trust NAC startup last year and integrated the technology into its Mist portfolio in a bid to compete against Cisco’s Identity Services Engine and Aruba Networks’ Clearpass NACs.

The new NAC solution allows Juniper to natively integrate NAC into its networking and security services, including its wired, wireless and SD-WAN portfolio, as well as Mist artificial intelligence (AI) and Connected Security services, Jeff Aaron, VP of enterprise marketing at Juniper Networks, told SDxCentral.

Aaron said Mist AI enables the NAC service to have insights into experiences and the nature of network issues. It also simplifies configuration, troubleshooting and operations with automation.

Additionally, Juniper continues to work with third-party NAC vendors including Cisco and Aruba, Gilby said. But compared to those solutions, the cloud-native and AI-driven NAC is simpler to deploy and roll out, offers a better user experience, finds problems faster, and allows partners and sales teams to spin the service up from the cloud, he touted.

Juniper has been steadily integrating AI capabilities into its networking portfolio since at least 2019 when the company acquired Mist AI for $405 million. And now the vendor has expanded it with ChatGPT.

Mist AI started out as a wireless local area network vendor, then it built its Marvis AI engine. Mist’s AI functionality won Juniper over, and now Mist serves as the heart of Juniper’s AI-based enterprise strategy, with Marvis as its brain.

To enhance Marvis’s conversational interface (CI) capabilities, the vendor is leveraging OpenAI’s ChatGPT API to bring even more human-like conversational capabilities, particularly with respect to documentation and support issues.

“LLM [large language model] enhances a VNA’s ability to accurately generate natural language text, particularly in more complex and nuanced scenarios. This combination improves customer experiences by providing easy-to-understand answers that reduce time spent looking for specific details,” Maribel Lopez, founder and principal analyst at Lopez Research LLC, said in a statement.

Aaron claims Juniper is one of the first networking vendors to start integrating ChatGPT and LLM into its model. “LLM doesn't replace the need for VNA like Marvis, because LLM isn't good for real-time queries like what is wrong with my Wi-Fi right now. But it's great for historical queries like what does the blinking light on an LED mean … so we moved our knowledge base into LLM.”

Juniper created its own private data store, using its own knowledge and token for the ChatGPT integration to make sure “it's not being polluted by what's out there” and offers accurate responses.

The capability is focused on helping users effectively mine the documentation from Juniper. It hasn’t fed customer data into the model, Gilby said. “I think there's definitely a level of concern from us from customers about feeding data in so we want to be really sensitive to that.”

“It's definitely kind of more generic and unidirectional. We've turned off the ability to change the LLM and it's only pulling data,” Aaron echoed.