3 September 2026LANTEK Computers
Manufacturing has a genuinely different problem to an office business, and most IT advice does not acknowledge it.
In an office, if a machine is running software that can no longer be updated, you replace the machine. On a plant floor, that machine is attached to equipment that cost a great deal of money, was commissioned with a specific software version, and is supported by a vendor whose position is that changing anything voids the arrangement. Replacing it is a capital project, not a Tuesday.
So the advice to keep everything patched, which is correct everywhere else, does not survive contact with a plant. Something else is needed.
The two networks problem
Most plants we see have one flat network. The office computers, the ERP, the wireless, the CCTV, the machine control PCs and the equipment itself are all on the same network, able to reach each other.
This happened for understandable reasons. The plant network was built when the machines needed to talk to the ERP for job data, somebody ran a cable, and it worked. Nobody designed it; it accumulated.
The consequence is that anything reaching any part of it reaches all of it. An office laptop that picks something up in an email is on the same network as the control PC for a production line. And the control PC is the machine that cannot be patched, cannot run modern security software, and in many cases cannot even be rebooted outside a maintenance window.
That is the actual risk in manufacturing, and it is architectural rather than about products.
Separating them, without stopping production
The principle is that the plant network and the office network should be distinct, with a controlled and minimal path between them. Not air-gapped, because the machines genuinely do need job data. Controlled.
What that means in practice, at the level worth discussing with your provider:
Establish what actually needs to talk to what. Almost always far less than the current arrangement allows. A control PC typically needs to reach one server for job data and nothing else. It does not need the internet, it does not need mail, and it does not need to be reachable from a laptop in reception.
Do it in the order that carries no production risk. Observe first, document what traffic actually exists, then restrict. Segmenting a plant network by guessing which flows matter is how a line stops.
Start with the direction that is safe. Stopping the plant network from reaching the internet is usually low risk and high value, because a control PC has no business browsing anywhere. Stopping the office from reaching the plant needs more care, because somebody’s reporting tool may depend on it.
Accept that the machine cannot be patched, and compensate. If a control PC has to stay on an old version, the answer is that it becomes something that can only reach one destination, cannot be reached from general workstations, and has no route out. The old software is then a much smaller problem, because reaching it requires already being somewhere specific.
Get the vendor’s position in writing. Machine vendors often say nothing may be changed. What they usually mean is that the control software and the machine’s own configuration may not change. Network restrictions around it are frequently acceptable, and sometimes recommended. Ask specifically, in writing, and put the answer in the maintenance file.
The other findings, consistently
Remote access for the machine vendor, permanently open. Commissioned years ago so the vendor could support the equipment, still enabled, often with credentials shared across their customers, sometimes reachable directly from the internet. This is one of the most serious things we find in plants and it is almost never on anybody’s list.
The fix is not to remove it, because you may genuinely need it. It is to make it on-request rather than permanent, and to know when it is used.
Nobody knows what the machine PCs are. No inventory, no record of versions, no idea which are still supported. You cannot protect an estate you have not listed, and this list is worth building even before anything is changed.
The ERP is a single point of failure with no tested restore. On a plant, the ERP going down stops dispatch, which stops revenue in a way an office outage does not. Test the restore, and time it.
Wireless coverage on the floor was never designed. Access points added where there was a socket. Handheld scanners drop, and the workaround becomes writing things down and entering them later, which is a data quality problem masquerading as a network problem.
Hard stops from power. A plant loses power like everybody else, and the machines and their control PCs take the same abrupt shutdowns. What load shedding does to your data applies with extra force where a controller is mid cycle.
Why this is worth doing before anything happens
The reason to address it is not compliance. It is that a plant’s tolerance for downtime is different in kind. An office can work on laptops for a day. A production line that stops has a queue behind it, contractual delivery commitments in front of it, and in some processes material that spoils.
And the recovery is harder, because you cannot simply restore a machine controller from a backup the way you would a file server. Frequently it means the vendor, on a plane or at least on a schedule.
Which means prevention is worth more here than in most sectors, and prevention is mostly the architecture rather than the products.
Where to start
Three things, in order, none of which touch production.
One, list every machine on the plant floor with a computer attached to it: what it is, what software version, is it supported, who supports it. Two, find out whether any of them can reach the internet, and whether any vendor remote access is permanently open. Three, get a documented answer from each machine vendor about what network restrictions they will accept.
Those three produce the plan. None of them require a maintenance window.
What we cover is on Manufacturing and Industry and Engineering and Construction, with the underlying work on IT Infrastructure. If you have a flat network carrying both the office and the floor, that is the conversation to have, and it starts with observing rather than changing anything.