A weekly routine for the one-person IT team of a 20-PC company

In a company with twenty computers, "IT" is usually one person whose real job is something else. They help accounting, fight the printer, look after the phone system, sometimes the camera at the front door too. Nothing gets checked on a schedule because there is no time to make a schedule. Things get looked at when they break.

Call it firefighting; nobody enjoys it. The annoying part is that some of those fires could have been seen weeks ahead: a disk filling up, a computer not updated for months, a license with two weeks left. Catching those takes about an hour a week. The list below splits that hour into five small pieces. Don't try to do it all in one sitting; you'll drop it by week four.

The times are rough. Yours may differ, which is fine.

What the week looks like

Monday, 15 minutes: what's on fire?

Start the week by answering that.

• On the panel home page, are there offline or long-silent devices? One quiet day is normal; the machine is off or the user is on leave. Three quiet days is a question mark.
• Is any device flagged? Old antivirus signature, BitLocker off, a disk above 90 percent, a new local administrator account. If a flag shows up, deal with that device this week. A flag you leave "for later" becomes ten flags in two weeks, and then nobody looks at any of them.
• Any pending commands? If something you sent remotely never completed, find out why. Usually the machine was off at the time. Sometimes it wasn't.

On the panel this is one screen.

Tuesday, 10 minutes: updates

Look at the machines with many pending updates and the ones that haven't restarted in a long time. Most Windows updates have actually downloaded; they're just waiting for a restart, which is why the list says "not installed".

A practical approach: pick the same day every week and tell people in advance. "Please restart your computer Tuesday afternoon" causes far less friction than forcing a restart from the panel. When you really need to, you can still send one; you pick the countdown, and the user sees a warning on screen.

Wednesday, 10 minutes: backups

This item is on the list on purpose, and EndpointGuard doesn't do it for you. You have to look yourself: date of the last backup, its size, any errors.

Once a month, go one step further and actually restore a single file. A backup you can't restore isn't a backup. We're writing this because it's the item most often missing from checklists like this one; everyone knows they take backups, few know they can restore them.

Thursday, 10 minutes: people and permissions

• Did anyone join or leave? Who takes over the leaver's computer, accounts, and any licenses tied to them?
• If you mark a person as "left" in the panel's personnel list, the devices assigned to them are released. The record isn't deleted, so "who was using this machine" still has an answer in the history.
• Has someone been added to the local administrators group? The device history shows it. If there's an account you don't recognise, look at that device that day, not tomorrow.

Friday, 10 minutes: what changed

Look back over the week.

• In the device histories, was any new software installed this week? If something unexpected appeared, who installed it and why?
• If you use IP Pinger, look at the event list. Did the gateway or the server drop out this week? If nobody complained you may not have noticed; and nobody complaining doesn't mean nothing happened.
• Do any licenses end within the next 30 days? Renewal talks with a vendor can take weeks.

Friday has one purpose: no surprises on Monday.

Once a month, 30 minutes

On top of the weekly pieces:

• Look at disk usage. The flag fires at 90 percent; a monthly glance at which disks are heading there buys you a few weeks.
• Compare the inventory with your purchasing records. Does the number of licenses you think you own match the number installed?
• Remove or tag retired devices. Old records make the list noisy, and a noisy list doesn't get read.
• Restore a backup for real (see Wednesday).

What you don't watch

A checklist should also know what it leaves out.

Don't monitor everything. A hundred indicators for twenty computers is too many; you end up reading none of them. Four flags, one event list, one license list. That much gets looked at.

If an alert never makes anyone act, switch it off. If something says "attention" five times a week and you say "ignore it" every time, you'll ignore the sixth one too, even when it's real.

One more thing: don't try to write every detail down somewhere. What goes into a spreadsheet is correct on the day it's typed and wrong a month later. A good inventory holds that information for you and tells you when it changes.

When you go on holiday

The real weakness of a one-person IT department is what happens when that person is away. The nice thing about a short list is that you can hand it to someone else.

You can create a user in the panel with view-only access. Before you leave, give that account to someone you trust at the company; ask them to open the Dashboard on Monday morning and call you if a device is flagged. No technical knowledge needed; as long as the answer to "is there a flag?" is on the screen, it works. That account can't send commands or change settings. It only looks.

Total: about an hour

Five small blocks and half an hour a month. The first weeks run longer, because that's when everything that piled up comes out. After the fourth week it mostly ends with "no flagged devices". At that point you've become someone who does this on a schedule, instead of putting out fires.

You can do the important parts by hand without any panel, but a weekly round across twenty computers by hand rarely lasts; it gets dropped within two months. The first five devices are free; try it for a week.