When drafting is automated, performance is no longer measured by what lawyers produce. It is measured by the quality of the decisions they make.

The billable hour was never a measure of expertise. It was a measure of time. For a long time, that was close enough.

Volume and speed became proxies for value. How many contracts a lawyer drafted, how quickly documents were produced, how efficiently revisions were completed. Billable hours, the total time logged and billed to clients, became the defining metric of productivity and worth.

Automation changes that calculus entirely. And for firms willing to follow that logic through, it changes it for the better.

What automation displaces

When drafting is automated, the routine is handled by the system. Templates, clause logic, and standardised workflows ensure consistent output at speed. Expertise no longer shows up in producing a document. It shows up in the decisions that shape what the system produces.

Most senior lawyers would acknowledge, if pressed, that the drafting itself was rarely the hard part. The hard part was knowing what the document needed to do, for this client, in this context, against this counterparty. Automation doesn't displace that judgment. It finally gives it room.

Where human judgment becomes visible

The human-in-the-loop moments in legal drafting are the points where that insight is not just useful but irreplaceable.

Designing the template requires a lawyer to make active choices about risk appetite, fallback positions, and negotiating strategy before a single document is generated. That is not administrative work. It is the distillation of institutional expertise into a system that will govern hundreds of future drafts.

Interpretation remains human territory. Understanding the nuance of a regulatory constraint, reading the commercial context behind a client's instructions, recognising when a standard clause creates unusual exposure in a specific deal: these are judgment calls that depend on experience and cannot be codified in advance.

Strategic intervention is where the most consequential decisions live. Knowing when a template needs to be adapted, when an exception requires bespoke drafting, when the deal has moved in a direction the standard form was never designed to handle. That is the work that distinguishes a capable lawyer from an automated process, and it is the work that automation makes more visible by clearing everything else away.

On HITL as a temporary safeguard

It can be argued that human-in-the-loop is simply a bridge, a necessary precaution until AI matures enough to guarantee both quality and consistency without human oversight. That framing misunderstands the argument.

Automation is not a replacement for expertise. It is the mechanism through which expertise scales. The question has never been whether AI will eventually produce accurate documents. It is whether accuracy alone is sufficient. A contract is not good because it is technically correct. It is good because it protects the right interests, enables the deal to move, and minimises the conditions for future dispute. Those are judgment calls. They require context, experience, and accountability that sits with a person, not a system.

Human-in-the-loop is not a temporary safeguard. It is the design principle that makes automation trustworthy.

What this means for how lawyers develop

If performance shifts from volume to the quality of interventions, the way junior lawyers are trained must shift with it.

This goes beyond teaching them to use AI tools. Automation gives junior lawyers something that routine drafting never did, access to the substance of legal work earlier in their careers. Time that was previously spent on mechanical drafting becomes time spent learning from playbooks, tracking which clauses are moving to market standard, identifying the positions that are hardest to hold in negotiation, and reviewing contracts across a portfolio rather than in isolation.

That kind of exposure builds the pattern recognition that distinguishes genuinely skilled lawyers. It develops judgment faster, and more deliberately, than grinding through first drafts ever could.

Redefining what performance means

The deeper implication of automation is not operational. It is professional.

For generations, legal performance was legible because it was visible. Hours logged, documents produced, revisions completed. Automation makes that ledger largely obsolete. What replaces it is harder to count but easier to defend: the quality of the judgment embedded in the system, the soundness of the decisions made at the points where humans intervene, the institutional standards that hold across every document the firm produces at scale.

The more productive question is not how to reduce human involvement in drafting. It is how to ensure that the involvement that remains is as skilled, deliberate, and well-placed as possible.

That is a different question. And it leads to a different kind of practice.