The Thin Client Is Dead: Your Workstations Just Became Critical Again

For about fifteen years, the office PC didn't really matter.

That sounds strange coming from a company that supports office PCs for a living, but it's true. Once businesses moved their email, files, and applications to the cloud, the computer on the desk became little more than a window. If it died, we swapped in another one, the employee logged back into their web apps, and everyone was working again within the hour. Nothing important lived on the machine itself.

Businesses noticed this, and they drew a reasonable conclusion: "Our data isn't on that workstation, so we don't need to back it up." For a long time, they were right. The old disaster recovery playbook — the one built around protecting local servers — quietly shrank down to a single question: is the cloud provider's backup good enough? Most small businesses shrugged, said yes, and moved on.

That era is ending, and most businesses haven't noticed yet.

What changed

AI changed it. Specifically, AI that writes working software on request.

Today an office employee can describe a tedious task to an AI assistant — "pull these numbers out of the invoices that come in by email and put them in the spreadsheet" — and get back a working script. No programmer, no software purchase, no IT ticket. It works, so they keep using it. Then they build another one. Then a scheduled task that runs every morning. Then a little database that keeps track of what's already been processed.

Six months later, that "thin client" on the desk isn't a window anymore. It's a bespoke automation machine running business processes that exist nowhere else — not in the cloud, not in any documentation, and often not fully in the head of the person who asked the AI to build them. They described what they wanted; the AI wrote how it works.

Here's the uncomfortable part: nobody decided this machine was important. It just became important, one helpful little script at a time.

The new failure story

The old failure story was simple. A PC dies, the employee is annoyed for an hour, work resumes on a spare.

The new failure story goes like this. A PC dies. The replacement arrives, the employee logs into the cloud apps, and everything seems fine — until Thursday, when someone asks why no invoices have gone out this week. The script that handled them is gone. The list of which invoices were already processed is gone. The log that would show what it actually did last month is gone. And the employee who set it up left in the spring.

The business didn't just lose a computer. It lost a process — and the memory of how that process worked.

Recognizing what the machine has become

The instinct here might be to blame the employee, or to ban the automation. That would be a mistake. These automations are genuinely valuable — they're why a five-person office can now do what used to take eight. The problem isn't that they exist. The problem is that the machines running them are still being treated like disposable thin clients.

Our view is old-fashioned on purpose: when a machine matters and you can't fully account for everything on it, you protect the whole machine. That means full image backups — a complete copy of the computer exactly as it is, every setting and script and quirk included — taken automatically, stored safely, and actually tested by restoring them. Not because it's clever, but because it's proven. It's the same standard businesses used to apply to their servers, applied to machines that have quietly become servers.

There's more to the full picture — finding these automations in the first place, protecting the logs and databases they create, and planning for the parts of recovery no backup can cover. We'll take those on one at a time in the rest of this series.

The question to ask this week

You don't need an IT background to start. Just ask around your office: "Is there anything running on your computer that does part of your job automatically?"

If the answer is yes — and increasingly, it is — then somewhere in your building there's a machine that matters far more than anyone has admitted, protected like a machine that doesn't.

← Back to all articles