HPE Aruba provides GenAI search for enterprise Wi-Fi users

  • HPE is using GenAI to improve the search function for its enterprise Wi-Fi users

  • The GenAI will be trained on HPE’s treasure trove of data from its GreenLake platform

  • Wi-Fi seems to be one of the first places where telcos are tapping GenAI

Hewlett Packard Enterprise (HPE) is going to use generative artificial intelligence (GenAI) in its Aruba Wi-Fi portfolio so users can search for answers to their questions about the portfolio.

User: Hey Aruba, why isn’t my company’s Wi-Fi working?
GenAI: Try unplugging and replugging your router.

The GenAI will be integrated into Aruba’s Networking Central, which is its cloud-native network management software. Networking Central does such things as provisioning and configuring Wi-Fi networks and orchestrating their traffic.

Alan Ni, senior director of edge marketing for Aruba, said the GenAI initiative has the goal of providing users of Networking Central with a far superior search function. He said AI search has been a tool within Networking Central for over two years. But until now, it’s been based off advanced natural language processing (NLP). 

“People ask questions, and we provide specific search results,” said Ni.

The legacy AI search function has already gotten quite a bit of usage, receiving over three million requests over the past two years of its existence from the likes of network IT operators, architects and engineers.

But now Aruba is taking it to the next level by tapping into its previous NLP AI tools — and it’s going to use the data from HPE’s giant GreenLake data lake where all these learnings have been stored. GreenLake has been around for about a decade. Aruba’s data, which includes telemetry from more than one billion unique customer endpoints, is included in GreenLake.


AWS Telecom GM sees less need for central data lakes with gen AI

To create the GenAI-based search function, Aruba will use multiple large language models (LLMs). 

“The LLMs are more effective and a better engine than NLP,” said Ni. “They’re advancing conversational models.” The LLMs will be trained on specific Aruba jargon, and they understand dialects and have a better grasp of what people are asking.

The three million questions that people have already asked of Aruba’s NLP-based search feature will be a good foundation for training the new LLMs.

The security question

Security is always an issue when customer data is being used. In this case, customer data will be used to train the LLMs.

Customers are very concerned that none of their personal or corporate identifiable data leave GreenLake, according to Li. She said, they wouldn’t want even the names of their Wi-Fi sites or the quantity of their sites to get in the wrong hands, and call logs are very sensitive by nature.

“The key way we achieve privacy is we’re using our own models and our own data to train. It’s self-contained and hosted on our platform,” said Ni. “We don’t have live APIs that facilitate customer information from leaving or entering that GreenLake platform.” 

As Fierce covered before, Wi-Fi seems to be the first area in telecom that’s tapping into the GenAI craze. Earlier this month, Nile, a startup headquartered in San Jose, Calif., said it was using AI and machine learning to offer a better Wi-Fi paradigm for enterprises.