The IT Department Taught Everyone Else to Work Without It. Now What?

There is a version of this story that gets told as a crisis. IT is dying. AI is taking over. The department that used to hold the keys is being bypassed, defunded, and quietly made irrelevant by the same business teams it spent two decades serving. I have heard that version in boardrooms, at conferences, in the LinkedIn posts of people who enjoy a good disruption narrative. It is not wrong exactly. It is just not the interesting part. The interesting part — the part most organisations are not yet structurally honest about — is that IT did this to itself. Not through failure. Through fifteen years of doing its job well. Every platform it rolled out, every tool it democratised, every system it handed to the business to run independently moved the organisation one step closer to the day the business stopped needing a gatekeeper. That day is not coming. For most organisations, it already arrived. The question is not what happens to IT. The question is whether IT figures out what it actually is now — before the next incident makes the answer for it. Gartner projects that 40% of enterprise applications will have AI agents embedded in them by end of 2026. Most of those agents are being deployed by business teams, not IT.

The business teams have caught up. A finance analyst who spent three years learning Salesforce, then Power BI, then Workday, is not waiting for a ticket queue to tell her what's possible. She is configuring AI agents now — pulling pipeline data, flagging anomalies, drafting board summaries — with the same self-sufficiency she brought to every tool before it. The operations manager who ran his own Power Automate flows in 2022 is running multi-step agents in 2025. They didn't need permission then. They don't need it now.

The IT department is not disappearing. But the version of it that derived organisational weight from being the sole interpreter between business need and technical execution — that version is already gone in the organisations paying attention, and going fast in the ones that aren't.

This piece is a clear-eyed read of what is actually happening inside organisations where AI agents are being adopted — not in the press releases, but in the operating model. It explains why the traditional IT department structure is losing its logic faster than most CIOs are comfortable admitting, what business teams are genuinely taking on and where they are struggling, and what the three structural shifts are that separate organisations navigating this well from those accumulating invisible risk. The reader will leave with a way of thinking about IT's future role that is neither the inflated "AI transformation hub" narrative nor the uncomfortable "IT is dead" provocation — but the more precise and actionable truth that sits between them.

Shift Three — Become the Function the Organisation Can Blame When Trust Breaks Down

This is the hardest shift because it requires IT to accept an identity that has no glamour in it. Not the builder. Not the innovator. Not the transformation lead. The guardian.

When a business team's agent exposes customer data it shouldn't have touched, someone needs to be accountable. When an AI-generated decision gets challenged in an audit and nobody can explain how it was made, someone needs to own that gap. When three different agents across three different business units are pulling from three different versions of the same data and producing contradictory outputs, someone needs to have seen that coming.

That someone is IT. Not because IT deployed those agents — it probably didn't. But because IT is the function responsible for the conditions under which any of that should have been possible in the first place.

The language most IT departments use to describe their evolving role is so broad it means nothing. "Enabling digital transformation." "Partnering with the business on AI." "Moving from build to enablement." These phrases sound like strategy. They are the absence of strategy.

The IT function that survives this transition will be defined by three things specifically:

Security and access architecture — who is accountable for ensuring that agents deployed across the organisation cannot touch data they have no business touching, cannot be exploited from the outside, and can be shut down cleanly if something goes wrong. As AI agent deployment scales, every agent is a potential exposure point. This is not optional and it is not a project. It is a standing responsibility.

Integration and data integrity — the reason most AI agent deployments fail at scale is not the agent. It is the data underneath it. Incomplete records. Conflicting sources. Systems that were never designed to be read by an autonomous process making consequential decisions. Someone has to own the plumbing — the standards, the pipelines, the data contracts — that determine whether the agents the business builds can actually be trusted. That is IT's job.

Security and access architecture — when the business deploys an agent that makes a decision affecting a customer, a claim, a financial transaction, or a compliance obligation, the organisation needs to be able to explain what happened and why. Audit trails. Explainability frameworks. Escalation protocols for when an agent hits a situation it was not designed for. The business team built the agent. IT builds the infrastructure that makes it defensible.

Business teams measure their own outcomes. That is as it should be — they own the agents, the workflows, and the results those workflows produce. IT's accountability is narrower and more specific: whether the infrastructure those outcomes depend on is trustworthy, auditable, and safe enough to stand behind when something goes wrong.

The diagnostic question: can your IT team point to a named owner for each of those three accountabilities today — not a committee, not a shared responsibility, but a specific person with a defined mandate? If not, the function has not yet made the shift.