By Adam Bihari, Vice President, Product
Physical security teams have never had more data about their devices, and rarely have they felt less in control of them. Fleets are growing into the tens or even hundreds of thousands, automation is doing work people used to do by hand, and AI agents are starting to ask for the keys.
In that world, three questions decide whether a platform actually helps you:
Here's where I stand on all three.
Most control planes let you see the problem. They don't let you fix it. People have multiple tools with multiple insights, but if the issues can't be fixed with those same tools, they're just expensive methods of knowing what's wrong. Everyone sees the same dashboard. No one is empowered to act. Data flows up; actions don't flow down.
My advice: Don't get trapped behind a pane of glass. Make sure remediation is part of the package. Data is great, but it's not the whole story. The system should have action and ownership baked into the design from the root up.
The last mile of control doesn't end at the console. A security operations lead needs to run remediation from the tools they already work in, not just watch it happen somewhere else. That means the system has to be built open from the start - APIs, standard interfaces like MCP, room for custom actions - so that action is available to every owner who needs it, whether that owner is a person, a script, or an agent.
To serve enterprise customers, every action has to clear three tests.
Authorized: Who is allowed to run it, and within what permission boundary?
Auditable: Is there an immutable record of what happened, and do I know where that record lives?
Accountable: Which system enforces that boundary, and can it answer for what happened when something goes wrong?
Skip any one of those three and blast radius stops being a design question and starts being an incident report.
Here's the part most buyers aren't thinking about yet: scale raises the stakes on every action. A person with a browser makes one mistake at a time. A script makes ten thousand identical mistakes in ninety seconds.
If I take an action on one device and it fails - it's not great, of course, but the damage is contained. If I take an action across 10,000 devices, I have to have the capability to approve it and make sure it happens in the right maintenance window. Sequencing is key here - operations need to happen in the right order, because devices don't live in isolation. Rollback matters too, especially for blast radius. Fleets are heterogeneous, and undoing a change is rarely as uniform as making it.
There's also a quieter risk: trust calibration. When automation works, operators stop reviewing it. The failure mode shifts from missed problems to unreviewed automation. Operators need to keep an eye on automated operations and make sure they go as planned. That's not a limitation of automation, it's part of running it responsibly. Humans still need to be in the loop.
None of this is theoretical anymore. Fleets are already at the size where these questions decide outcomes, and the population of actors is about to grow - what scripts do today, agents will do tomorrow. Who owns the guardrails when an agent takes the action is a big enough question that it deserves a post of its own, and it's one I'll come back to.
For now, the takeaway is simpler. The platforms that earn trust in this space won't be the ones with the best dashboards. They'll be the ones that treated action, ownership, and scale as first-class design decisions from day one. That's the standard we hold ourselves to - and the standard I would advise any buyer to expect from their vendors.