When drafting is commoditised, judgement is the product

For most of its history, a large part of legal work has been the production of documents. Contracts, advice notes, letters before action, board minutes, leases: the craft of turning a client’s situation into careful, defensible prose has been both what lawyers do and, through the billable hour, a good deal of how they are paid. That settlement is now under pressure, because the production of competent first-draft prose is precisely what generative artificial intelligence does cheaply and at scale.

It would be a mistake to overstate how far this has gone. As the courts have been at pains to point out, these tools fabricate, and a draft that reads well can be wrong in ways that matter enormously. But it would be a bigger mistake to assume the pressure is imaginary. The direction of travel is that the mechanical component of drafting, assembling standard clauses, producing a serviceable first version, summarising a long document, is becoming abundant and therefore cheap. When something becomes abundant, its price falls, and work priced by the hour spent producing it comes under strain.

The instinctive defensive move, insisting that a machine cannot really draft as well as a good lawyer, is true but beside the point. Clients have rarely been paying for prose as such. They have been paying for the judgement embedded in it: the decision about which risks to accept and which to resist, the read on what a counterparty will actually do, the knowledge of where a clause has caused trouble before, the willingness to be accountable for advice. Those are not tasks a language model performs, because they are not really language tasks at all. They are judgement, and judgement is the thing that does not commoditise.

The implication for firms is uncomfortable but clarifying. If the drafting layer is thinning in value, then billing models built on the hours it consumed will erode, and the firms that thrive will be those that can charge for outcomes and judgement rather than for time at a keyboard. That, in turn, changes what junior lawyers are for. The traditional apprenticeship ran through exactly the document-production work that is now most easily automated, which raises a genuine and unresolved question about how the next generation acquires the judgement that the senior work depends on. You cannot supervise a machine’s draft well if you never learned to write one yourself.

None of this is a prediction that lawyers are obsolete; the profession has absorbed word processors, email and document automation before without disappearing. It is an argument about where the value sits. The parts of the job that were always about knowledge, responsibility and judgement are becoming the whole of the job’s defensible core. Firms that understand that, and price and train accordingly, will be fine. Those that keep selling the drafting and hoping no one notices what it now costs to produce are the ones with a problem.