A ticket came in last spring with a one-line description: "it's doing the thing again."
I knew exactly what she meant. That is the part that should worry you.
I have spent most of my working life on the receiving end of messages like that, and I want to open with something that sounds like the setup for a complaint but isn't. The people who file the hardest requests to work are not careless. They are busy, describing a machine they did not build in words nobody ever taught them, while something they need right now is broken. Under those conditions I file terrible tickets too. I opened one with our ISP in March that said, in full, "slow again."
So none of what follows is a scolding. It is the contents of my head after years of decoding, plus the half of the problem that belongs to my side of the desk, which is the larger half.

A K-12 team supporting a few thousand devices told me the single hardest recurring problem with submitted requests had nothing to do with writing quality. It was categories. Faculty could not consistently pick a route for a damage report, because the line between a cracked screen protector and a cracked screen is invisible to anyone who is not holding the repair invoice. Same device, same crack, two queues, two very different costs.
The downstream effect was brutal. Free-form requests without a clear category could sit unprocessed for weeks. Not because anyone was lazy, but because people do not pay enough attention to queues, and a request nobody has classified belongs to nobody in particular.
Weeks. On a broken screen. In a building where the person was sitting forty feet away.
I want to be precise about whose problem that is. If your intake asks a colleague to correctly guess your internal taxonomy, you have handed a data-entry job to someone who does not have your schema. That team's fix was to replace the free-text box with categorized buttons: lost charger, broken screen, missing device. Picking "lost charger" now generates two linked records automatically, one to bill it and one to ship the replacement.
What changed there was the help desk workflow. Nobody was trained in anything. Which brings me to my least comfortable opinion here: most of what we call user error at intake is us charging people a fee for our own missing structure.
Here is the observation that rearranged how I read a queue. Someone put it to me almost exactly this way: your users, customers, and clients don't always tell you what the problem is, they tell you everything else around it.
You cannot unhear that. "I was on the phone with the vendor and then the file wouldn't send and Sarah said hers worked yesterday and I have a meeting at two." Four facts, none of them the problem, all of them adjacent to it. The problem is genuinely in there. Decoding it is my job, which means the burden here is shared rather than assigned: they describe, we translate, and neither half does anything useful alone.
Two habits do most of the work.
Never take a self-diagnosis at face value. Not because people are unreliable, but because a diagnosis is a conclusion and what you want is the observation underneath it. Did they actually quit the application, or close the window and leave it running? Is the thing charged? Has it been restarted since the last time it worked? I have closed more IT helpdesk issues with those three questions than with every diagnostic tool I own, and I still ask them apologetically, because nobody enjoys being asked whether it is plugged in.
Ask for a screenshot with markup, or a short screen recording. Circles and arrows. Ten seconds of video beats two careful paragraphs, and the reason is not laziness on either side. What someone is describing might not be what is actually happening. The word "crashed" covers a hang, a spinning cursor, a force quit, an app that closed cleanly, and an error dialog with the answer printed on it in plain English. A screenshot ends that ambiguity instantly.
Some complaints are load-bearing myths. Wi-Fi is the worst offender.
One team traced a persistent dead zone to a copy room, where the stacked wall of paper turned out to be, in their words, darn near radio opaque, degrading signal for everyone working near the big multi-function printer. Another found a network switch sitting physically on top of a microwave oven, which produced chaos concentrated around lunchtime and looked, from the queue, like a scheduling mystery. My favorite is the school next to an airport whose Wi-Fi died in sync with a runway-approach radar sweep. Fully diagnosable. Completely unfixable, short of the school relocating.
"The network is slow" is even better, because so often it is physical. A chewed cable. A badly punched-down jack. Wire pairs untwisted too far back from the connector, because whoever terminated it stripped more than necessary since that makes crimping easier. Modern Ethernet runs at frequencies close to radio, so a nicked conductor produces symptoms that look exactly like a configuration bug and will send you chasing your tail for a full day. The cruelest version I know of is a cable with two bad pairs that were not the pairs used for negotiation. Link came up green and healthy. No data ever flowed.
And then there is the quintessential support call, which will outlive all of us: the person who keeps running over the Ethernet cable under her desk.
Two more deserve to be famous.
"My PDFs open in the wrong app." Filed after an application migration, usually with an apology for being fussy about it. The actual cause is that installing a new application never changes existing per-user default-handler associations, so every file someone double-clicks keeps launching the old app until something explicitly rewrites the mapping. Nobody did anything wrong. The operating system simply does not behave the way every reasonable person assumes it does.
A print job that lights up like it's going to print, and then nothing comes out. That exact description is worth more to me than any log I will pull. It reliably means a driver incompatibility rather than a network or queue problem, and it saves both of us forty minutes of restarting spoolers.

Copy this. Paste it into your next ticket. Sixty seconds, and it changes what happens next.
What I was doing: exporting the Q3 deck to PDF from PowerPoint.
What I expected: a PDF on my desktop.
What happened instead: progress bar reaches roughly 80%, PowerPoint greys out, no file appears. Screenshot attached with the stuck bar circled.
When: today, 9:15am and again at 9:40am.
Reproducible: yes, three times, same deck. A different deck exported fine.
Already tried: actually quit PowerPoint (cmd-Q, rather than closing the window), restarted the Mac, exported to a different folder.
Urgency: client presentation at 2pm today, so I need it before then. Otherwise not on fire.
That last line matters more than people realize. Not everything is urgent, and "urgent" with no reason attached gets deprioritized against a request that explains itself. When somebody writes "client at 2pm," they have handed me the ability to sequence my entire afternoon correctly. When every request is flagged high priority, I am sorting by guess.
A short please-don't list, offered with real affection:
Any article like this that only lists what employees should do differently is smug and deserves to be closed unread. So here is the reciprocal half, and I think it is the more actionable one.
Canned acknowledgments. An auto-reply that describes a queue transmits exactly one fact: you are somebody's problem later. "Hey, saw this, I'm on it" carries identical information about timing and completely different information about whether a human being is involved.
Silent fixes. Repairing something without explaining what happened leaves a person exactly as confident as they were before, so they file the same request in March and feel sheepish about it. Thirty seconds of "here is what went wrong, here is what to try first next time" is the cheapest training your organization will ever run.
The caught-out reflex. When somebody admits they installed something odd or clicked a link they shouldn't have, the correct response is a version of "what you tried makes sense, here's a way that'll probably go smoother next time." Anything colder buys you silence later, and silence is expensive.
Our vocabulary. Nobody outside our team knows what a profile, a payload, or a queue is. I once watched a genuinely kind engineer tell a nurse that her MDM record had failed to reconcile.
Sorting. Our help desk tools hand us dashboards, ticket numbers, statuses, and views sorted every possible way except by person. Someone with three open requests and two weeks of waiting looks completely fine in every view we have built. Once a week I sort by requester instead. It is uncomfortable, and it is the most useful ten minutes in my week.
I now read the pre-apology as a measurement.
"Sorry to bug you." "This is probably a dumb question." "I know you're busy." For years I found those charming and replied "no bother at all!" and felt like a good colleague about it. I was wrong about what they were. They are a reading on how expensive it feels to ask us for help, and nobody apologizes to a process that feels cheap.
That cost never arrives as a complaint. It arrives two other ways. People delay asking, so we hear about problems later than we should, secondhand, usually in a meeting where we look unprepared. And they solve things themselves: one team discovered staff had been downloading whatever free tool they needed and hoping it was fine, because asking IT felt like it carried a price.
Nobody filed that. It showed up as an unmanaged application on a laptop nine months later.
The best reframe I have heard on this came from an admin who stopped explaining tickets as procedure and started explaining them as memory. A record means the pattern gets tracked. The third time a machine does the same strange thing, the record is what turns "huh, that again" into "that is the same fault three times, we are replacing the board." Filing it makes your own next problem faster to fix, and your colleague's too.
Her line: once people understand the process is there to help them, it stops feeling like a barrier.
Which is the real test, and it tests us rather than them. If knowing how to submit a support ticket properly requires internal knowledge nobody published, structured intake buttons will fix more of it than any instructions I write, and building those buttons is our work rather than a puzzle for everyone else to solve around.
We are the only department that gets to see everybody's worst five minutes of the day. The screen that went black before the deadline. The call that dropped mid-sentence with a customer who was already annoyed. People are rarely at their most articulate in that moment, and I do not think they should have to be.
The best thing anyone has ever said about a team I ran had nothing to do with uptime. She said we were easy to ask.
Foqal builds conversational ticketing for IT teams in Slack and Microsoft Teams, with structured intake that does the categorizing so your colleagues don't have to guess.
Get the latest insights on IT operations, AI, and workplace productivity delivered to your inbox.
See how Foqal can help your team deliver faster, smarter support.
Start Free TrialA small IT team can turn chaotic, untracked scripts into a reliable, auditable system by moving to Git‑backed version control, protected branches, and automated CI pipelines that sync and safely delete changes. This approach eliminates guesswork, adds security, and ensures every edit is traceable.
An honest split between the AI IT support tasks that genuinely work now (classification, log triage, knowledge base answers) and the ones that fail badly, with the failure data and real incidents.
One org cut two hours of IT prep per device to zero and about 30 minutes of user setup. Then the hard problems moved into people's homes: bandwidth, missing cables, and old laptops that never came back.