Somewhere in your firm, right now, someone on the knowledge team is trying to build a standard. Maybe not even a grand programme. One document. The kind of agreement the firm produces a hundred times a year, where everyone would be better off starting from a single version they trust. 

It's hard work. It means tracking down the version each team uses now, the one buried in a recent matter rather than the one that has sat on the intranet for years. It means pulling a dozen executed contracts up alongside each other and working out, clause by clause, which position the firm genuinely stands behind. Some of what's in those contracts is the firm's real view. Some of it was whatever a partner conceded at eight on a Friday to get a deal signed. Telling them apart is most of the job. Then comes the hardest part: getting several partners, each quietly certain their own version is the right one, to agree on one. Weeks of toil, not a word of it approved yet.

And here's the part that really hurts. By the time the standard clears review, the ground has already shifted. A market-norm has changed. A partner has just won a sharper clause on a live matter. Writing on boilerplate this year, The Lawyer made the point that terms have to keep adapting as circumstances change. The standard behind those terms moves the same way. You finish the job, and it's already a version behind.

So firms do the rational thing and build only a few. Not out of laziness... Each standard costs months, so it goes only on the documents where the pain is worst: the highest-volume agreement, the one that caused a real problem last year. Everything else, most of what the firm produces, never gets a standard at all. Not because those documents don't matter, but because no one has the months. So it runs on whatever the last lawyer happened to open.

Even complete, a precedent only helps the lawyer who opens it. To get it working on every matter, you have to automate. That's what we've spent years building: automation that carries your best work across the whole firm without losing a word of it. But automation only pays back if people use it. That's where most systems fall: make them clumsy and lawyers route around them. So the hard part was never the automation, it was making it easy enough that a busy lawyer reaches for it by default. That is the problem we set out to solve. And even then, most firms are only part-way there: what they've standardised and automated so far is a fraction of what they produce.

For years that gap was an internal frustration and little more. The knowledge team knew. Everyone else billed. What's changed is that it now sits under every firm's AI plans. It's the quiet reason so many of them disappoint.

Legal IT Insider put it plainly this year: AI can only be as effective as the data beneath it. Point the best model in the world at a firm with no agreed standard for the document in front of it, and it has nothing of the firm's own to reach for. So it reaches for the market average instead. The result looks competent. It's also indistinguishable from what the firm across the road produced from the same model.

Feeding it the firm's own files helps, but less than you'd hope. In most firms, that data is fragmented, inconsistent and short on governance. So the model sounds like the firm, but it can't tell the firm's best work from its most ordinary, or the position the firm stands behind from the one a partner conceded on a Friday to close a deal. Without a standard marking out the best, it averages everything: the firm's middle, not its edge. AI was never going to be as good as it could be, because it never had the one thing it needed: the firm's best work, set as the standard.

That gap costs money even before AI. In research reported by Artificial Lawyer, IDC put the drag from fragmented data at around 11% of revenue lost or delayed, with more than eight in ten in-house teams saying admin routinely crowds out the work that matters. The standard nobody had time to build is billed to the firm every year, whether or not anyone writes it down.

So flip it around. If the missing standard is what holds a firm back, then having one, kept current and trusted, is the edge. Above the Law, drawing on research from Stanford's CodeX centre, calls the first draft the "power move": whoever owns the starting point owns the terms of everything that follows. Set your best work as the standard and you own that starting point on every matter, not just the few someone thought to prepare for.

And unlike a finished document, a living standard doesn't sit still and lose value. It does the opposite. Every deal the firm closes makes the next one sharper, because every improvement goes back in. A standard you keep is worth more every year. A standard you build once is worth less by the month. That's what compounds. It shows up where it counts: sharper positions, better outcomes, the loyalty that keeps good clients coming back. So the work was always valuable. What multiplies it is never letting it stand still. Capture the best version. Put it to use on real matters. Feed every improvement back in. Done that way, the same loop that keeps your standard improving is what lets AI draft as your firm rather than the market. Use the models for reach and the open questions. Draw the answer you put your name to from work you've already approved. Both, not either.

Building a firm's standard and keeping it alive has always been some of the most important work a firm does. What's changed is that it's now also the most strategic. Build it once and it decays. Keep it moving and it compounds, until the standard drafting your fiftieth deal is sharper than the one that drafted your first. That's the whole shift: not a document a firm finishes, but a living asset it never stops improving.

Eliot Benzecrit

Co-founder, Avvoka