Do you think that you can use AI in IT work without losing your job? I do. I am 52, and over the past year, AI has done more work for me than several of the juniors I used to onboard.

I also work less than I did, and I have more time for myself. The tasks that used to eat my evenings, I close before lunch. And let me be clear and honest here: I use AI not to work more; I use AI to work less. I am not like a Meta manager who wrote an email to employees that since they have access to AI, they should deliver much more work...

Everyone is asking whether AI will take their job in IT. That is the wrong question. The useful one is how to use AI in IT work without losing your job - because both outcomes are happening right now, and which one happens to you is largely a matter of what you do next.

I am not an AI enthusiast selling a course. I run ChatGPT, Claude, and Perplexity every day on real client work. Here is where it helps and where it falls over.

What I use AI for in IT work

Some tasks have genuinely disappeared, and good, because they were the ones stealing the most time:

• Writing PowerShell from scratch. An hour once. Two minutes for a working skeleton now.
• Reading long Microsoft documentation. I paste it, ask the question, get the answer with a source link.
• First drafts. Client emails, audit reports, incident write-ups.
• Translating unfamiliar logs and errors. The ones I have never seen before and would have spent forty minutes googling.

That is the boring, repetitive job. Work nobody enjoyed, and work that used to keep me at the desk long after everyone sensible had stopped.

Do the arithmetic. Two hours a day is ten hours a week - an entire extra working day. The only real question is whether you give that day to yourself or stuff more of the same work into it.

What it will not touch

Now the part that decides whether your job is safe.

Responsibility. If you misconfigure Conditional Access and lock an entire company out on Monday morning, you answer that phone. Not the model. Remember, AI will tell you later: "Oops, my mistake," but you will be responsible in the end...

Client context. AI does not know that this particular CEO clicks every phishing email, or that the accountant holds permissions she should never have been granted, or that one server in this environment always throws the same alert and it is fine.

Decisions and trust. A client is not paying for a script. They are paying for someone who takes on the risk and says: we will do it this way, and I own the outcome.

The useful way to think about it: AI is a brilliant junior. Fast, well-read, never sleeps. And occasionally says something completely wrong in a confident voice, with no awareness that it is wrong.

Somebody has to be the senior who catches that. That is your new job description.

What a real day with AI looks like

AI helps productivity" is an empty phrase, so here are three concrete cases from my own work.

Case one: the forgotten SharePoint function. I needed to declare records in SharePoint Online via PowerShell. I know PowerShell well enough, but writing a script around a function that God and man have essentially forgotten was a genuine challenge. Who still uses records in SharePoint?

Under normal conditions, that is hours, possibly days, including trawling forums. Working with an AI agent, I built a working solution in a single session - one that then saved me dozens of hours, because removing record declarations from hundreds of files by hand is not a thing a person should do.

Case two: Microsoft Graph scripting. I describe the problem, I get a ready script. Then - and this is the whole point - I read it line by line. I know what Get-MgUser does. I know what I am looking for. I fix two places where the model used a deprecated command.

Forty minutes of work done in eight. Those 32 minutes do not go into another task. They go into a walk, or into recording a video. But I only reclaimed them because I know what I am reading.

Case three: alert triage. A dozen Defender and Sentinel alerts waiting in the morning used to mean manual digging on each one. Now I paste the raw alert in and get a plain-language translation plus a first hypothesis about where to look. Triage that took half a morning happens over the first coffee.

I still decide what is a false positive and what needs immediate action, because the model does not know this environment.

One distinction worth making: I am not talking about an off-the-shelf chatbot. I am talking about building your own setup with memory files that learn, day by day, how you work and what you expect. That takes time to build. It is worth building.

The pattern is always the same. AI does the first 70 percent and does it fast. The last 30 percent - judgment, context, decision - is done by a human who understands the subject.

That is exactly why the reclaimed hours are safe. Nobody takes them from me as long as I am the one who understands. If you do not understand the part AI got wrong, you paste, you run, you pray. That is not working in IT. That is roulette.

Five steps to use AI in IT work without losing your job

1. Stop fighting it. Pick one task today. Take the most boring thing on this week's list and do it with AI. Measure how long it took before and how long it takes now. Write the difference down. That number is your reclaimed time.

2. Learn the fundamentals harder, not softer. This sounds backward, and it is the whole game. The more AI generates, the more it matters that you can judge whether the output is any good. Networking, identity, security fundamentals - that is your edge over the colleague who only copies prompts.

3. Watch where the time goes. Worst-case scenario: AI saves you two hours, and you immediately push two more hours of work into the gap. Then your calendar wins, and you do not. Deliberately return part of that time to yourself.

4. Specialize where mistakes are expensive. Security, compliance, architecture. The higher the cost of being wrong, the less any company will trust a machine alone, and the less replaceable your judgment becomes.

5. Learn to ask well. Describing a problem precisely is a real skill now. The clearer your thinking, the better the answer and the less time you spend correcting the machine. It is the same skill good engineers always had - the audience just changed.

The rule has not actually changed

Twenty-five years ago I knew nothing about IT. I studied journalism. I bought a thick book on Exchange administration, learned it almost by heart, went to the interview, and got the job.

My advantage then was that I learned faster than other people were willing to.

Today it is the same advantage. AI did not change the rules of the game; it raised the tempo. The winner is not the person who knows the most, because knowledge is one prompt away now. The winner is the person who learns fastest, owns the outcome, and spends the freed-up time deliberately.

I had that edge in 2000. I have it in 2026. It is available to you too.

Bottom line

AI will not take your job in IT. Used well, it hands you back your time. Ignored, it hands your work to someone who is using it.

Pick one task this week and measure the difference. And if you are wondering whether this changes the calculus on entering the field at all, I looked at that in cybersecurity career change after 40.

📧 I write weekly about M365 security for SMBs - join the newsletter on LinkedIn → My linkedIn newsletter