The door knows whether that worker's training is up to date.

Almost every access control system answers one question: can this credential open this door at this time? Command Centre is built to answer one more: does this person also meet what is needed to go in there? The working-at-height course in date, the electrical risk training, the medical check, the contractor's accreditation. If that has expired, the door does not open and nobody has to remember.

That changes the kind of problem it solves. It is not designed mainly to keep an intruder out —it does that— but to keep out somebody who should not be there even though they work for the company. It is the difference between a security system and one that also holds up health and safety.

The other thing that places it is the perimeter: the same system that runs the doors runs the intruder alarm and the fence, including the electrified fence, with the same console and the same log. You find it in industrial plants, data centres, water and energy infrastructure, ports and sites with a lot of fencing: places where the system starts long before the first door.

Its own server with a database, one console that serves both for administering and for operating, and a line of controllers that is its own: it does not drive other manufacturers' electronics as the backbone of the architecture. That has two sides. In favour: the whole thing is designed as one, the supervision of the line and the reader is fine-grained, the encryption runs end to end and new features appear across the whole system. Against: if your wall is already full of controllers from another family, they are not reused here.

The field cabling follows an idea of its own: a single cable carries power and data to the readers and the modules, which simplifies the installation a great deal and at the same time ties it to its own components. Third-party readers can be hung off the classic protocol, and it is done, but then you give up part of the supervision this system was chosen for. It is a legitimate decision and it has to be taken knowing what is being given up.

For the credential it takes encrypted cards, the phone and third-party biometrics, and it has a self-service part that gets little use: the employee or the contractor submit their paperwork and the manager approves it from the phone. Outwards it has a well-documented programmatic integration route, which is what lets the training system, the personnel system or the contractor management system talk to the access control without anybody copying lists by hand.

Making access conditional on an accreditation being in date is excellent and it is useless if nobody feeds in the dates. Somebody has to tell the system that that worker renewed the course on Tuesday. If it is typed in by hand, by the third month it is out of date and the usual patch appears: a permanent permission so as not to fight with the door. The real project is not fitting readers: it is deciding where those dates come from and connecting it to the system that has them.

The second is the count of who is inside each zone. It is one of the most useful things it does —evacuations, nobody alone in a confined space, how many people are in a dangerous area— and it only works if you also read on the way out. That means a reader in both directions, with its cost and its building work, and discipline in using it. A system that counts badly is worse than one that does not count, because it gets used in the belief that it counts well.

It fits when going into somewhere depends on a condition and not only on a name: training in date, a work permit, an accredited contractor, spaces with a risk. It fits when there is a perimeter to watch and you want it in the same system as the doors. And it fits where the documentation of who went in and with what authorisation has to survive an inspection.

It does not fit as well in an office building that only wants its doors in order and has a wall full of controllers from another family: there you pay to change all the electronics to use a small part of what this platform can do. And it does not fit where nobody is going to keep the accreditations up: the feature that justifies the price sits switched off.

The day to day is run by the client, no argument: people, credentials, schedules, zones, visitors, reports. And here there is one more level worth using because it is the one that holds the model up: that the manager of each area approves or refuses the access for their zone, even from the phone, and that the accreditations are kept by whoever really manages them, which is hardly ever security. A system where security has to keep track of the training courses of the whole workforce does not last six months.

What the client should not touch is the structure: creating zones, changing the matrix of conditions, touching the perimeter and the intruder detection, changing what the doors do in a fire, moving up a version. Not out of possessiveness: because those pieces are interlinked and a change in one zone can alter the presence count or the automatic arming of another.

For them to run it, more than usual has to be handed over, because the model is richer: the matrix of conditions written down, who is responsible for each zone with their full name, the procedure for what to do when an accreditation expires and somebody has to go in anyway —because it is going to happen—, and training for three roles: whoever manages people, whoever approves access for their area and whoever operates it. Without the procedure for the exception, the system ends up full of permanent permissions granted in a hurry, and that hollows it out.

Plainly: if what is in place is electronics from another family, the controllers are changed. There is no shortcut. What is almost always kept is the data cabling back to the cabinets and the builder's work, which is a large part of the cost, and it can be done building by building with the two systems living side by side while the transition lasts.

The readers are where the decision is. Third-party ones can be kept on the classic protocol, and for many internal doors that is reasonable. On the doors that matter it is worth fitting the platform's own reader and cabling, because the supervision and the end-to-end encryption are exactly what you came for. A mixed installation decided properly is worth more than a uniform one decided out of inertia.