As AI gets better at generating configs, scripts, and fixes, a lot of people in IT are quietly worried about what's left for them. I think the answer becomes clear the moment something goes wrong in production at two in the morning, and it comes down to a single word the machine can never supply: accountability.
An AI can suggest a change. It can propose a firewall rule, write a deployment script, or recommend a fix for an incident, and increasingly it can do all of that impressively well. What it cannot do is be responsible for the outcome. When that suggested change takes down a critical system, the model doesn't get paged. It doesn't understand what the outage costs the business, it can't weigh whether this particular risk is acceptable for this particular company, and it can't stand behind the decision when someone asks who signed off. Accountability requires a someone. The machine is a something.
This is why I don't think the future of IT is about competing with AI on generating solutions. It's about owning the decisions that surround them. Knowing what to build and what to refuse. Understanding the specific context of your organization — its risk tolerance, its regulatory constraints, its history of what has burned it before — none of which the model has access to. Judging whether a suggestion is safe here, in this system, under this load, for these users. And carrying the responsibility when a call has to be made.
The routine work will absolutely shift toward the machine, and that's fine. Let it draft the boilerplate, propose the first-pass config, surface the likely cause of an incident. That was never the valuable part. The valuable part was always the judgment layered on top: deciding, in context, whether the suggestion should actually be trusted, and being the one who answers for it either way.
So if you're wondering where you fit as these tools improve, don't try to out-generate them. Become the person who owns the outcome. Use AI to move faster on the routine, and pour your real attention into the judgment and the accountability, because that's the part that doesn't transfer to a model. A machine can suggest all day. Someone still has to be responsible for what actually ships, and that someone is the job.
– Serguey Shinder
