A Project Manager once told me, in a tone that suggested this was obvious, that IT existed to facilitate the needs of Production and resolve their tickets.

He was not being dismissive. He was being sincere. That is what made it worth thinking about.

What That View Assumes

The helpdesk model of IT has a coherent internal logic. Users have problems. IT fixes them. The faster the fix, the better IT is performing. Metrics are clean. Accountability is simple. Everyone understands the transaction.

It also assumes that the problems users can see are the only problems that matter.

They are not.

The problems users cannot see, the ones that accumulate quietly in the background of every studio that treats IT as a support function, are the ones that eventually surface as compliance violations, security incidents, unplanned outages, and the specific kind of expensive chaos that arrives when nobody was governing the infrastructure because everyone assumed someone else was.

What AI Just Made Worse

Generative AI tools have accelerated something that was already happening. Production teams move fast. They adopt tools that solve immediate problems. They do not, by default, think about where the data goes, who owns the outputs, what the licensing terms say about commercial use, or whether the API key being used belongs to a personal account or a company account with no audit trail.

That is not a criticism. That is the natural behavior of a team optimizing for throughput.

But every AI tool adopted without IT involvement is a governance gap. It is a potential data residency violation. It is an unlicensed asset waiting to become a legal dispute. It is credentials stored somewhere that will never appear in an offboarding checklist. It is a model trained on studio IP under terms nobody read.

None of these failures announce themselves at the ticket queue.

The Two Studios

Consider two studios at the same scale, with the same headcount, building the same kind of game.

In the first studio, IT is a helpdesk. It resolves access requests, replaces hardware, and maintains the network. When Production wants to integrate a new AI concepting tool, they sign up, share the credentials across the team in a group chat, and get to work. IT finds out when something breaks. Security finds out during an audit, if there is one. Legal finds out when there is a dispute about generated assets used in a shipped product.

In the second studio, IT is infrastructure. When Production wants to integrate the same AI concepting tool, IT is in the room before the contract is signed. The licensing terms are reviewed. The data processing agreement is assessed. The tool is integrated into the identity management system so access is provisioned and deprovisioned with the rest of the user lifecycle. The outputs are logged. The cost is tracked against a known budget line. When an artist leaves, their access to the tool leaves with them.

Both studios shipped a game. One of them has a defensible audit trail. The other has a group chat with a password in it.

What IT Actually Governs in a Modern Studio

When AI is part of the production pipeline, IT is no longer managing network drives and build servers as its primary function. The scope is different now.

It includes the infrastructure that makes AI workloads run, including GPU provisioning, cloud architecture, storage at scale, latency between systems that need to communicate in real time. It includes the licensing and vendor governance for every AI service the studio touches, which in a mid-sized studio in 2026 is likely more services than anyone has formally counted. It includes the security posture of those services, meaning how credentials are managed, how outputs are stored, how access is controlled when someone moves teams or leaves the company. It includes the compliance and audit readiness that determines whether the studio can defend its use of AI-generated assets if that use is ever challenged.

None of that fits in a ticket queue.

The Contractor Math

There is a version of the helpdesk model that its proponents sometimes use to justify it: if IT is not doing infrastructure, then contractors can do infrastructure when it is needed.

This is worth examining directly.

A studio that outsources infrastructure governance to contractors does not save money. It defers cost and loses continuity. The contractor who builds the cloud environment does not maintain it. The contractor who handles the security audit does not monitor it between audits. The contractor who reviews the AI vendor contracts does not track the renewals or the usage. Each function, separated from the others and handed to an external party on demand, costs more and produces less than a single internal function that owns the full scope.

The helpdesk model does not eliminate the need for infrastructure governance. It just means the studio pays more for it, gets less of it, and has nobody internally who understands the whole picture when something goes wrong.

The Actual Question

The Project Manager's comment was not wrong about what IT does in a helpdesk model. He was describing the reality of studios that have made that choice, intentionally or by default.

The question worth asking is not whether IT should resolve support tickets. It should. The question is whether that is the ceiling or the floor.

In a studio where AI is now part of the production pipeline, treating IT as a helpdesk is not a lean organizational choice. It is a governance gap with a timer on it.

The timer runs until the first audit, the first breach, the first legal challenge to an AI-generated asset, or the first time a critical vendor integration fails and nobody inside the building knows how it was built.

Whichever comes first.