In-house is not a smaller law firm.

There is a persistent assumption that a legal department is a law firm with fewer people and no invoices. It leads directly to bad advice, because the economics run in precisely the opposite direction and almost everything else follows from that.

A firm sells hours and must generate demand for them. A department absorbs demand it cannot decline, against a budget set by something other than that demand. Efficiency in a firm reduces revenue. Efficiency in a department is the only lever that exists. Once you see the inversion, several familiar departmental pathologies stop looking like management failures and start looking like structural consequences.

The absence of billing pressure, for instance, means most departments have worse data about their own throughput than a similarly sized firm. Nobody was ever required to record where the time went, so nobody knows. A department attempting to make a capacity case to a CFO frequently discovers it cannot describe its own workload in numbers, which is not a discipline problem so much as an artefact of never having had a reason to.

The absence of a gate on incoming work matters even more. A firm can decline an engagement. A department receives requests from colleagues who can escalate, through whatever channel they happen to know, at whatever level of specification they feel like providing. This is why the bottleneck in a legal department is almost always intake rather than production - and why the instinctive first automation, contract drafting, is usually the wrong first move. The lawyer is not slow at drafting. The lawyer is spending a third of the day working out what each request actually is.

The compensating advantage is that the business case is trivial. A firm that automates has to decide what to do about the revenue it just deleted. A department that automates immediately serves a queue it was already failing to serve. There is no cannibalisation problem, no compensation model to protect, and no partner vote. The unmet demand is sitting right there.

What is hard instead is everything downstream of the corporate context. Procurement, security review, data residency, enterprise architecture, and the fact that the legal function does not control its own technology stack. This is why most departments run AI Adoption rather than a full rebuild - the primitives can all be named and systematised without asking anyone's permission, but the automation layer installs inside constraints that are not up for negotiation. Recognising that early is the difference between a programme that ships and one that spends two quarters waiting for an approval it never scoped.