IT is both the industry building AI and one of the industries being most reshaped by it — a genuinely unusual position to be in.
Last reviewed: October 2026
On a typical engineering team, AI shows up less as one big tool and more as a layer sitting inside tools people already use. It's the autocomplete inside the code editor suggesting the next ten lines. It's a chat assistant explaining an unfamiliar error message or an old codebase nobody documented. It's a first-pass summary of a long incident postmortem so the on-call engineer doesn't have to read every Slack message from 3 a.m. None of this removes the need for engineers who understand the system — it removes a lot of the typing and searching that used to eat up their day.
Teams that get the most value tend to treat AI output the same way they'd treat a capable but very new junior teammate: useful for a fast first draft, not yet trusted for architecture decisions or anything touching production data without review. Teams that get burned are usually the ones that skip that review step because the AI-generated code "looked fine."
Software and IT are unusually good environments for AI because many outputs can be checked: code can run, configurations can be validated and logs can be compared. I still would not paste secrets, credentials or private customer data into an unapproved model.
A practical test: Start with a task that has an objective check—writing a small test, explaining a log entry or drafting documentation. The faster the verification loop, the safer it is to learn where the model helps and where it guesses.
This is a template, not a tested result. Replace the brackets with your own details:
For the habits behind these checks, see Fact-Checking AI Output and Privacy & Data When Using AI.
The clearest near-term effect is on the volume of routine, boilerplate coding work — not on the need for engineers who can design systems, debug tricky issues, and make architecture trade-offs. The job is shifting toward more review, judgment, and system-level thinking, and away from typing every line by hand.
It depends entirely on the tool's data policy. Some coding assistants train on your code by default unless you opt out; others explicitly don't. Always check the vendor's data-handling terms before connecting an AI tool to a private or proprietary repository.
Reported gains vary widely by task type — routine, well-documented coding sees the biggest speedups, while novel or highly domain-specific work sees much smaller gains. Treat any single "X% faster" claim as a starting estimate, not a guarantee for your specific codebase.