What to watch for as enterprises move from AI experimentation to autonomous operations
SiliconANGLE Zeus Kerravala
Enterprises are moving from AI demos to systems that can act on their own. The hard part isn’t smarter models; it’s trust, security and control.
Based on reporting by SiliconANGLE, Zeus Kerravala — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
For the past three years, enterprise AI has mostly been about what models can say. The next phase is about what those systems are allowed to do. That’s a much messier question, and better models by themselves won’t solve it.
That shift is the backdrop for Wednesday’s Bloomberg Tech Event in Boston, where the crowd is heavy with the people who actually have to live with agentic AI: CIOs from Mercedes-Benz North America and elsewhere, security leaders from Booking Holdings and Carnival Corp., Estée Lauder’s head of data, analytics and AI, and HPE’s James Robertson. That mix matters. Vendors can sell autonomy all day. Operators decide when it gets turned on.
The biggest change is simple to say and hard to handle: AI has moved from insight to action. Most companies already have plenty of recommendations. The real test is whether a system can take a step without human approval every single time. Drafting a bad email is one thing. Changing a network policy, rerouting traffic or quarantining a device is another. Get that wrong and the business can go offline.
That’s why uptime is no longer the right scorecard. HPE’s Sujai Hajela put it bluntly at Networking Investor Day: “Up is not the same as good.” A network can stay online and still ruin a video call or slow an AI workload. The more useful measure is business outcome — fewer incidents, less manual toil, faster service delivery. ServiceNow’s Sajeev Nair described a setup where years of point products were creating more telemetry fragmentation, and after consolidation a job that used to swallow 2,000 to 3,000 hours a year now takes fewer than 60.
Security will probably set the pace. AI agents need access to systems, data and each other, which creates a new population of nonhuman identities. At the same time, attackers have AI too. Anthropic’s Claude Mythos has helped vetted organizations identify more than 10,000 high- or critical-severity vulnerabilities, and older unsupported network gear looks increasingly exposed. HPE’s David Hughes argued that the network can act as a sensor and contain threats itself, extending zero trust to agents and devices. That’s not a side issue. It’s the gatekeeper.
The blockers, though, are still human and operational: fragmented data, legacy infrastructure, unclear accountability, skills gaps and trust. HPE says its self-driving network is supposed to show operators what it recommends and why before acting. Good. That’s the minimum. The enterprises that get to autonomous operations first won’t be the ones with the flashiest model. They’ll be the ones that clean up the plumbing, set the rules and actually mean them.
My take — AI-written commentary, not fact-checked reporting
The industry loves calling this “autonomy,” but most of the work is still just discipline with a nicer name. The real winners won’t be the loudest AI vendors; they’ll be the CIOs and CISOs willing to say no, patch old gear and limit what agents can touch. That’s boring, which is usually how enterprise progress actually happens.
Read more about this at: SiliconANGLE