A key goal for many C-suite tech leaders in the AI-powered world is to make the technology deliver. Figuring out how AI fits into their organization’s operations can include deciding what AI even means for a company.
“At the end of the day, we’re trying to automate things that typically a human would do. That’s the same thing with the car, right? I mean, would you call the self-driving car AI or fancy automation?” asked Bob Friday, chief AI officer at Hewlett Packard Enterprise (HPE).
Friday sat down at NRF 2026: Retail’s Big Show recently to share a little bit of how HPE uses AI, and how he has leveraged the technology as it has evolved.
HPE offers networking, servers and containerization software and services. Friday was a co-founder of Mist Systems, an AI-powered wireless LAN startup that was acquired by Juniper Networks, which was in turn acquired by HPE.
Where AI met real-world constraints
Mist — and his work in AI — got their start, Friday explained, when he tried to convince a big retailer to put a mobile experience app on its network.
“They told me they weren’t going to do it unless I could promise them I could stop controllers from crashing, innovate more than twice a year and guarantee a great user experience,” Friday said.
AI became part of making that happen. Friday said the retailer’s request was an early indicator of a paradigm shift from classic network support to cloud AIOps.
AI adoption isn’t just technical
That shift also had a cultural dimension, Friday said. It meant persuading IT support teams to let AI into the mix of day-to-day operations. Getting network support staffers to embrace that change — and surrender their Secure Shell protocol keys used for encrypted remote access — remains a challenge. “It’s like taking candy from the baby,” he said. “They did not want to give up those keys.”
Working through that AI-supported model for the retailer also taught Friday that AI is not a one-size-fits-all technology.
“When I started Mist, I always believed that natural language was going to be the next user interface,” Friday said.
Networking had already moved beyond command-line interfaces to dashboards and then to natural language, which Friday said is a logical step for pulling data out of databases.
The rise of ChatGPT and agentic AI opened up new possibilities. “That’s when [AI] became much more powerful around reasoning and generation,” he said.
Agentic AI also changed how developers approach their work because it is a nonlinear, nondeterministic programming language, Friday said. “It’s basically a new way to automate problems that have been hard in the past.”
For instance, AI could lead to networks that correct problems without human users at the wheel. Getting to that stage will require IT personnel to trust AI to address network issues.
“That’s very similar to your self-driving test. When you finally trust the car, you’d take the driver out of the seat and let it go,” Friday said.
Where AI needs limits
Use of AI can get ahead of itself, Friday cautioned. “If you have hundreds of APIs, you just can’t expose them all to the large language models … it will confuse the model,” he said. Hierarchy and organization remain important when introducing those APIs before LLMs can process what they are dealing with.
Why people stay in the loop
The human element, at least for now, also remains essential to AI’s use — as long as IT professionals keep their skills up-to-date. In the old days, programmers wrote code, tested and verified that the code performed as intended, Friday said. That linear process has changed with the advent of nonlinear AI.
Why developers can struggle with the shift
Friday said developers who have spent years programming in linear ways can struggle to make the shift to nonlinear AI. “They have a hard time with this new programming paradigm,” he said. “You’re better off to hire a kid out of school who’s not locked into programming a certain style, a certain way.”
