Is IT still a good career now that AI can write code?

This is now the most common question I get from parents and from students deciding what to do after school. It deserves a straight answer rather than reassurance.

The honest position

Yes, IT is still a good career. But the shape of it is changing, and some of the traditional entry points are genuinely harder than they were.

Anyone telling you nothing has changed is not paying attention. Anyone telling you the field is finished is also wrong, and usually selling engagement.

What AI is genuinely doing

It writes a lot of routine code. Boilerplate, standard functions, tests, conversions between formats. Work that used to occupy junior developers for hours is now minutes of generating and checking.

It is very good at explaining and at first drafts. Documentation, configuration files, scripts to automate a task.

It handles a growing share of tier-one support questions. Password resets, how-do-I questions, simple troubleshooting.

That is real, and it has already compressed some junior work.

What it is not doing

Deciding what to build. Translating a vague business problem into a technical requirement remains entirely human, and it is most of what senior people actually do.

Being accountable. When a system fails at 3am, somebody has to own it. An organisation cannot put an AI on the incident report.

Physical work. Racking hardware, running cable, replacing a failed drive, fixing a network in a building.

Anything requiring trust and context. Sitting with a frightened staff member. Judging whether a vendor is telling the truth. Knowing that the finance system must not be touched during month end because of what happened last year.

Diagnosis in a messy real environment. AI is good at problems that resemble problems it has seen. Real faults are frequently strange, involve three systems interacting badly, and depend on undocumented local history.

What is actually changing

The bottom rung is higher. If your value was writing simple code to a specification, that value has fallen. This is the genuinely hard part, and it is felt most by new graduates.

Verification is a skill now. Being able to look at generated output and know whether it is right is worth more than being able to produce it slowly.

Breadth is worth more. Somebody who understands networking, systems and security together can direct these tools usefully across all of it. Narrow specialists in one routine task are more exposed.

The infrastructure underneath is growing, not shrinking. All of this runs on enormous quantities of compute, storage and networking that somebody has to build and keep running. That work has increased.

Which parts I would head toward

Infrastructure, networking and systems. Physical, contextual, and hard to automate. The demand is growing because AI itself needs it.

Security. Attackers are using these tools too, the surface keeps expanding, and there has been a shortage for a decade.

Anything involving people. Support that requires trust, technical roles that translate between business and engineering, anything where somebody has to sit with a human being.

Working with the tools rather than against them. People who understand what these systems do, where they fail, and how to build with them safely are in demand right now and there are not many of them.

And what I would be more careful about

Pure junior coding on well-defined tasks. Not gone, but more competitive, and the expectations of a graduate are higher than they were.

Straightforward tier-one support. Increasingly handled before it reaches a person.

Neither is a reason to avoid the field. Both are reasons not to stop at the first rung.

What has not changed at all

The advice for somebody starting is almost identical to what it was five years ago, which is itself informative.

Understand the fundamentals. How computers, networks and operating systems actually work. These tools generate answers, and you need to know enough to tell when the answer is wrong.

Build things. Practical experience matters more now, not less, because generated knowledge is abundant and demonstrated capability is not.

Learn to diagnose. The single most durable skill in the field, and the one AI is furthest from replacing.

Learn to explain things to people who are not technical. Increasingly the differentiator.

For parents deciding whether to encourage it

If your child is genuinely interested in how things work, this remains one of the better fields available. It pays well, it is portable, there are many routes in that do not require university, and the demand has not gone away.

What has changed is that “did a course” is no longer sufficient on its own. What separates people now is having built something real and being able to talk about it.

That is the gap I try to fill, through practical tutoring and building a lab they keep. It is the same advice I would give whether or not AI existed. It has simply become more urgent.