
My AI can beat up your AI.
Doers buy legal tech. Builders make it. My theory is that the ones doing both, call them builder-doers, will always outclass the ones who only buy.
Every drafting and review platform sold to lawyers is built for the median matter at the median firm, and your deal isn’t the median anything, so there’s always a stretch between what it hands back and what the client needs. Buy only and you cover that stretch by hand, on every matter, forever. Build and you cover it once and it stays covered.
Every hour, Claude reads my inbox and turns what I owe people into Clio tasks. On document work it also works out how many hours the job takes, off a rate table built from my time entries, and writes it in before I’ve logged in. Nobody sold me that. If it’s technically feasible, you can have it, shaped however you want. A doer files a feature request and waits on somebody else’s roadmap.
I also can’t supervise a tool I’ve never opened. Building is how I learn where these models go wrong, and they go wrong confidently, in the voice of somebody who has read the entire file. If you’ve only ever typed into a box, you can’t calibrate it.
The reason I say always rather than for now is that the gap regenerates. Whatever I build, the platforms eventually ship some version, and by then I’m twenty problems further down the road. And that delta ever increases.
The work will also concentrate. Fewer lawyers doing more work, gravitating to whoever can best bend the tools to the matter.
Depending on how large that class of builder-doers is, there may not be enough work left for the doers.
Since reading product terms is a great deal of what I do, a warning to the doers and their buyers. Renewal windows are short, seat minimums ratchet, and getting your own work product back out is usually an afterthought. A three-year term locks you to somebody else’s roadmap through 2029 in a market that reprices about every quarter. It’s also three years you spend not learning to build while the people who are keep getting further ahead.
There’s a bill for building. Time I’m not billing anyone, and a fair share of what I build I throw out.
Give me some time. Soon there will be enough data points, and I’m looking forward to proving this out.
So, yeah. My AI can beat up your AI. Yours demos better.
Attorney Advertising. Prior results do not guarantee a similar outcome. Not legal advice.



