Governing AI, Not Just Adopting It: A Field Note From HPE Discover
Key Takeaways
Introduction
We spend a lot of time at vendor events. Across the partners we work with, we are looking for the same thing every time: not the flashiest demo, but the shifts that change how our clients should run their environment. HPE Discover was one of several stops for our team this season. We have sat with what we saw for a few weeks now, compared it against the other events on the calendar, and identified what was important to share with our audience.
A note on where we sit. Derive is a full-service integrator, not a single-vendor shop. HPE is one of many partners we build with, and our job is to analyze at what they share with the same clear eyes and translate what matters for the people who trust our guidance. We are comparing what we saw across the ecosystem, take this as a field note, not a sales pitch.
Here is what stood out.
The conversation moved from adopting AI to governing it.
The announcements worth caring about were not about doing more on autopilot. They were about control: who is allowed to act, what gets a human sign-off, and how you undo a bad decision. HPE put real structure behind this, a layered model for running AI agents in production. Verify the user, govern what the agent itself can do, and gate the actions that matter behind human approval, with the ability to roll an environment back to a clean state if an agent gets it wrong.
We saw the same theme everywhere our team went this season, convincing us this is more than just a marketing campaign. Agentic AI has moved out of the chat window and into production: agents that take an action on a trigger, hand off to one another, and keep a human in the loop to check the result rather than approve every step. The maturity question the whole market is now converging on is not whether an agent can run on its own, but which actions it should be trusted to take, and what happens when it is wrong. We expect every serious vendor to have answers on that from here forward.
Networking moved back to the center.
With Juniper now folded in alongside Aruba, HPE is positioning networking as the core of the platform, with self-driving networks spanning the edge, campus, data center, and AI factory, and intelligence shared across the two lines. In practice the pitch is managing the network in plain language, and managing it centrally from the cloud. The example that stuck with us: firmware across thousands of switches, where you set the policy, decide what the system can do on its own, and the rollout runs in the background. You authorize the intent; the network executes it. Worth saying plainly a few weeks on: this is the announced direction and the demo, and the real test is how it holds up in production over the coming quarters, which is exactly what we will be watching as clients put it to work.
This is not an HPE-only bet. Across the networking partners we track, the same two ideas keep showing up: cloud-based management as the default rather than the exception, and network automation treated as a first-class discipline instead of a side project. The vendors are racing to bring bleeding-edge capability to ordinary IT teams without requiring an army of certified specialists to operate it. That convergence is the story, and it is good for the clients who have to run these environments.
"What stood out to me on the partner side is the cross-platform continuity and the governance built in. It means we can bring this to clients on top of what they already run. Because we were comparing it against everything else in the market at the same time, we can be honest with a client about where it fits and where it doesn't. That is what turns an announcement into something deployable."
Justin Marcus, Director of Partner Alliance.
Security is moving into the fabric.
Inline inspection on the switch itself means the network becomes a place you enforce policy, not just move packets, and it happens without the overhead you would expect. Security built in, not bolted on, was a common thread across the events our team attended this season, not a single vendor's slogan. You still protect the edge and the endpoints. Now the fabric is a control point too.
Resilience you can count on.
The recovery story does double duty. It can undo a bad agent action, and it anchors cyber recovery built on replicated data rather than backups, so restoring is closer to flipping a switch than pulling from tape. When a real ransomware event can lock up data for weeks, that speed is the whole game. And the threat is getting more human. One case study shared at the event traced an attack that ran a target through several rounds of fake interviews before slipping a corrupt file onto a corporate machine. Technology alone does not catch a process that patient. Process and people must be in the loop.
The honest read.
Every major vendor is racing toward AI-native infrastructure right now, and the competitive lines are redrawing fast. The questions are changing, and the right answer depends entirely on what you already run, what you are regulated to protect, and where you are headed.
What it means if you are the one accountable for the environment.
The platforms are building the controls in: identity tiers, approval gates, rollback. But controls only matter if they are configured for how your environment actually works. Anyone can sell you a box or a model. The work is what happens next: standing it up inside a regulated environment, setting the guardrails, deciding what the AI is allowed to do on its own and what stays in human hands, and fitting it to how your people operate. AI does not remove judgment. It concentrates it. The question stops being can it run on its own and becomes what should it run on its own, and who decides.
Answering new questions like these is the work we have done for 25 years, across our full partner ecosystem, for clients in healthcare, financial services, and the public sector. We were on the floor at Discover with our engineers, reading it the way we read every partner: for what it changes for you.
If you are weighing what to hand to AI and what to keep in human hands, that is the conversation we have every day, across whatever you already run. Reach out and let's have it.