GSX 2026 Recap: AI Everywhere, and What It Takes to Put It to Work
Published on: September 30, 2026
Walking the floor at GSX 2026 in Atlanta two weeks ago, the biggest takeaway was a lot of familiar topics that have developed considerably this year. AI and the convergence of physical security and IT have been part of the conversation for several years. What changed was how each in its own way has moved from a sense of skepticism into stages of real progress and thoughtful application.
AI is no longer an analytics add-on shown in a corner of a booth; it is built into the cameras, access control, and platforms teams already buy. Teams want to use it, and they also recognize what comes with it: more complexity across fleets that were already difficult to configure and manage, and new requirements for governing what AI does. Both add to a challenge that has not gone away: aligning physical security with IT.
These takeaways from the show floor were heavily reinforced in our more in-depth conversations. Over three days we met with end users running physical security programs of every size, and spent time with dozens of customers and partners. Three topics came up in nearly every one of those conversations:
- AI in products. Teams want to use the AI capabilities manufacturers are shipping, and recognize that each one adds configuration and management complexity to fleets that are already diverse.
- AI governance. As agents become part of security operations, the question is shifting from what AI can do to how its work gets applied safely and consistently across the fleet.
- Physical security and IT alignment. Convergence is happening, but in deeper conversations most teams described alignment that is still hard to reach, and AI adds new pressure to it.
AI in products: valuable only when the fleet is managed well
The AI capabilities now arriving in physical security devices and platforms are impressive, and the teams we spoke with want to use them. These features share a dependency that rarely makes the product sheet: they only deliver value if the devices themselves are proactively maintained, both operationally and from a cybersecurity perspective. An analytic delivers nothing from a camera that is offline or misconfigured, many new capabilities require a minimum firmware version, and an expired certificate can break the secure connection that carries a device's data to the platform or leave it running with a known security gap.
AI raises the stakes on fleet management, because more of the operation now depends on every device working as intended.
New capabilities also add to a problem that already existed. Most enterprises run devices from many manufacturers, across many models, sites, and firmware versions. Each new feature is another setting to enable, validate, and maintain across that mix, and today much of that work is still manual and device by device. The result, as several attendees told us, is that capabilities they have already paid for go unused. Even device manufacturers themselves acknowledge that their innovation is outpacing end user application.
Getting value from AI in products requires a platform that can execute tasks at scale across every device and every new capability, regardless of manufacturer, and confirm the result.
AI governance: using agents is one thing, applying their work is another
As AI moves from flagging events to taking action, governance emerged as the clear follow-up question. When we showed teams how AI can now generate device operations from a plain-language description, and how our MCP server connects AI assistants directly to device data, the first reaction was interest. The second was a set of practical questions: What is the agent allowed to change? Who approves it? How do we know it worked, and where is the record?
Those questions point to the real challenge. Getting an agent to produce a configuration change or a remediation task is increasingly easy. Applying that work safely, consistently, and at scale across thousands of devices and multiple manufacturers is not. It requires policy that defines what can run automatically and what needs approval, execution that behaves the same way on every device it touches, validation that the change took effect, and an audit trail that physical security, IT, and cyber teams can all review.
Organizations adopting agents in security operations will need that governed execution layer between the agent and the fleet. Without it, AI either stays in pilot or introduces risk into the systems it was meant to improve.
Physical security and IT: the finger-pointing hasn’t gone away entirely
Everyone we spoke with agreed that physical security and IT are converging. Physical security devices run on the corporate network, and IT and cybersecurity teams have a legitimate stake in how they are secured. When conversations went deeper, most teams described the same gap. Physical security owns the devices and depends on them every day. IT and cyber set the policies those devices need to meet. But each team works from its own tools and its own view of the fleet, so questions like which devices are out of compliance, who is responsible for fixing them, and whether the fix was made rarely have a shared answer.
In meetings with prospective customers, this dynamic of accidental adversaries was an almost ever-present topic. One physical security professional equated it to the infamous Spiderman meme where 3 costumed characters are all pointing at one another.
Fortunately, we also spent a lot of time with some of our more advanced customers. Thos discussions highlighted how convergence works when all three teams work from one real-time source of truth for every connected device. Physical security keeps operational ownership and the ability to act across the fleet. IT and cyber get continuous visibility into device health and compliance, and evidence that policy is being met. Remediation happens on the device itself, not only as a finding passed from one team to another. That shared foundation is what turns convergence from an org chart into an operating model.
Looking ahead
The progression we saw at GSX points in one direction. AI will keep arriving in every layer of the physical security stack, and the teams that benefit will be those that can manage their fleets at scale, govern how AI acts on them, and give physical security, IT, and cyber a shared view of the same devices. That is the work we are focused on at SecuriThings.
If you’d like to learn more about how we’re helping in each of these areas, explore these links:
