Ask a lawyer how to argue a motion and there is precedent. Ask how to structure a deal and there is precedent. Ask how their own firm runs a conflicts check and the answer is: the way we have always done it, which is not the way anyone else does it, and no one is quite sure why.
Law is the most precedent-driven profession there is. Every argument stands on an earlier one. But that discipline stops at the edge of the practice. Cross into the business of law, into intake, conflicts, billing, matter setup, and collections, and you leave the world of shared authority for one where every firm invented its own answers, alone, and never compared notes.
The practice has a “how.” The business doesn't.
The practice of law is universal in a way that is easy to miss. Research works the same in Seattle as it does in Miami. Drafting follows conventions taught in every law school, tested on every bar exam, enforced by every court. The how of practicing law is standardized, documented, and public.
The business of law is the opposite. Its operations grew up ad hoc, and there is nothing to cite:
- No restatement of how to run a conflicts check.
- No casebook on client intake.
- No model order for matter setup or docketing.
- No treatise on billing and collections.
Instead there is a workflow bolted on in 1997, a spreadsheet someone built in 2009, and a “that is just how Denise does it” that outlived Denise.
You can watch it happen inside a single firm. We asked a litigation team that assembles court motions hundreds of times a month how they decide what goes where. The answer, from a partner: some attorneys “do it bespoke,” and others just “name the file in the folder.” Same task, same office, two methods, nothing written down. Now multiply that by every firm in the country.
Which is why the most common question at any legal-operations gathering is some version of “How do other firms handle conflict checks?” These are institutions whose entire method is looking up what others did before them, and on the basic operations of their own business there is nothing to look up. No precedent exists.
Why AI landed on the practice first
This explains something about the current wave of legal AI that the demos never do. AI went after research, drafting, and review not because that is where the value is, but because that is where the how is. A model can learn to draft because drafting is universal and documented: teach it once and it works for every firm on day one. AI is very good at standardized problems, and the practice of law is one.
The business of law offers no such foothold. There is no universal conflicts process to train against, because there is no universal conflicts process at all. Automating operations means first confronting the fact that the operations were never designed. They accreted.
Firms feel this the moment they try. One litigation practice spent the better part of a year with a vendor trying to automate just two of its document workflows before it stalled. As a partner put it, “We thought this would be easier than it is.” The tooling was fine. The workflow lived in a dozen people's heads, slightly differently in each.
The service is the business. The process was bolted on.
In a manufacturer, a bank, or a logistics company, the process is the business: engineered, documented, measured, refined, because process is the product. Law firms inverted this. The service is the business, the lawyer's judgment is the product, and the process was an afterthought, scaffolding no one meant to look at again.
The cost hides in plain sight, because the operational hour is not just unglamorous, it is unowned, so it quietly expands. Ask partners about time entry, the operational task closest to revenue, and you hear it: at one firm, two partners in the same practice admitted to being weeks behind on their own billing. The work that pays the firm is the work most likely to be done late, because the process around it was never built.
That inversion made sense when process was cheap and lawyers were the constraint. It stops making sense the moment technology can absorb operational work. The old catch was that you could not automate a process you could not describe, and most firms, asked to describe theirs, cannot. What has changed is that describing it is no longer a project the firm has to finish first.
The next precedent gets written now
Here is the reframe: the absence of precedent is not only the problem. It is the opening.
The mistake is to treat it as an audit, to freeze everything, get every process documented, and only then automate. That is the project that never finishes, and it is why most of these efforts stall. The process does not have to be written down before AI can run it. It can be captured from how the work is already done, turned into one standard way of doing it, and automated from there. The written-down version, the precedent, is the output, not the prerequisite.
This is also why buying more AI is not the answer on its own. We have sat with firms running a hundred of their own AI workflows whose billing was still done by hand, because the tooling sat on top of an operational layer no one had ever standardized. Tools accelerate a process. They do not create one, and they do not get better on their own.
That is the whole premise of Caddi. Not to touch the judgment you sell, but to take the operational work around it, the intake, the matter setup, the billing, the collections, and learn how your firm already does it. Caddi turns that into one standard way of running the task, automates it, and improves it over time with AI. You do not write the process down first. Caddi standardizes it, runs it, and makes it better, so your operations stop depending on who happens to be at the desk.