The wrong starting point.

The hardest part of bringing AI into a law firm is not choosing between Harvey, Legora, and Claude. It is that most firms treat that choice as the first question, when it is almost the last.

Each of these tools has real drawbacks. They cost money. They have limits. They take time to learn, and they change under you. Those are genuine problems, and weighing them is part of the work we do. But none of them is the problem.

The problem is the order. "Let's bring in Harvey" is a technology-first decision. It answers a question the firm has not yet asked. It assumes the firm already knows what its work is, how that work creates value, and where judgment can be assisted and where it cannot. Most firms have never written any of that down.

What starting with the technology produces

Start with the tool and you get a generic configuration of legal work, not a system that reflects your firm. Partners do not see their own practice in it. The economics that actually govern the firm go unexamined. Adoption stalls, and not because the model is weak: it stalls because nothing was built around the people who were supposed to use it.

This is the pattern behind most failed implementations. The technology works in the demo and dies in the firm, because the demo was never the hard part.

What starting with the need looks like

Starting with the need reverses the sequence. First you codify how the firm actually practices, and how it actually makes money. Then you decide what should be automated, what should be assisted, and what should be left alone. Only then do you choose and configure a tool, and by that point the choice is straightforward, because you know exactly what you are asking it to do.

It is also why the specific tool matters less than the order. The best model today may not be the best model next year, and the vendors will keep changing their limits and their prices. A firm that started with its own practice can change the tool underneath it. A firm that started with the tool has to start over.

Capability is no longer the constraint

Something has shifted in the last two years. For most of what a firm does, the question is no longer whether a tool can do it. It's whether you've decided what you want it to do. If you can describe the practice you want, you can now build toward it. That puts the burden back where it belongs: on judgment, not on software, and judgment is exactly the part a tool can't supply for you.

This is what a technology-first decision quietly gives up. Start with Harvey or Claude and you aren't picturing what your firm could become; you're accepting the version of legal work the product already assumes. You stop deciding and start complying. The firms that will pull ahead are the ones that do the imagining first, and treat the tool as the thing that carries it out.

The order is the strategy

Get the order right and the technology becomes a detail, an implementation question, solved late, with confidence. Get it wrong and no budget, no vendor support, and no model upgrade will save it. So before you ask which tool, ask the question the tool is supposed to answer. That question is your practice, and it belongs first.

← Back to insights