The order is fixed.

Most firms coming to us have already tried the third step. They bought a drafting tool, or a research tool, or an intake widget, and bolted it onto the practice as it exists. Six months later, the tool is either abandoned or ignored. The lawyer concludes AI is overhyped.

The lawyer is not wrong about the outcome. They are wrong about the cause. The tool did not fail because the tool was bad. It failed because it was asked to automate something that was never named, and never systematised. Automation is the last thing you do, not the first.

Name means: the primitives of your practice - matter, actions, time, team, evidence, research - get written down, in your firm's own language, with their edges drawn. Systematise means: those primitives are given shape as templates, playbooks, role charts and deadline structures, so the practice no longer relies on any one person's memory. Only then does automate mean anything.

There is a reason for the strictness of the order. An AI agent, unlike a junior lawyer, cannot compensate for the ambiguity of an unnamed system. It cannot 'figure out' what a matter is if you have not defined one. Give it clarity and it is genuinely useful. Give it fog and it produces more fog, faster.

The good news, if you are running an independent practice, is that this sequence is entirely within reach. A three-lawyer firm can complete the full pass in four to eight weeks. A three-hundred-partner BigLaw partnership cannot complete the same full rebuild on any honest timeline - it runs AI Adoption instead, because step one, name, requires deciding, and deciding requires an owner. Independent practices have owners. That is their edge.