Where to start with AI in a law firm is one thing. Not a system. One thing that is broken, fixed inside a week, by someone who does not need you to redesign your practice first. Sell a firm-wide rebuild as the entry product and you lose the buyer before the first call ends, because the buyer is already overwhelmed and you just confirmed it.
There is also a point where that stops being true. A threshold where I tell a firm to stop buying components from me and commission a build instead. I am going to name it in the next section, because a vendor who will not name it is not being straight with you.
At roughly half a dozen tools, stop buying components and commission a build
Once a firm is running about half a dozen separate tools that all came from the same source, piecemeal has stopped being the right answer. That is the line. Past it, the honest recommendation is to stop buying the cheap thing and pay for a build that knows about all of the jobs at once.
The reason is arithmetic, not sales. Each tool was scoped alone. Six of them overlap. Six of them ask for intake separately. Six of them hold a partial picture of the same matter and none of them holds the whole one, so somebody on staff becomes the integration layer. The cost of stitching six things together, in staff hours nobody is tracking, exceeds what one properly scoped system would have cost at the start.
I would rather say that at tool five than have a firm discover it at tool nine and conclude the whole category was a waste. It costs a couple of sales in the short run. It is the only version of this business worth running.
Where to start with AI in a law firm: one bite
Below that threshold, modular is not a compromise. It is the correct answer, and it is how the conversation should be framed. The framing that converts, from a strategy call:
How do you eat an elephant? A bite at a time. A lot of these people are overwhelmed by redoing their whole system, so it is a lot easier to say here is what you can do with this one thing.
Overwhelm is not a marketing problem you fix with a better deck. It is an accurate response to being told your entire way of working is wrong by someone who has never carried your calendar. One thing, working, this week, is the only pitch that respects the position the buyer is actually in.
Pick the thing that hurts weekly and has a clean edge. Financial affidavit intake. The first pass on opposing party disclosure. The correspondence you rewrite from scratch every time because the last version is in an email thread somewhere. Small, bounded, verifiable in one matter.
Two family law firms in one day, both saying the same sentence
Two on the calendar in a single day. Both inbound, neither prospected. Both saying they see and hear about AI and law firm automation everywhere, have no idea what it actually is, do not know where to start, and want help. Neither one asked for transformation. Neither one used the word platform.
That is the real state of the market. It is not a state in which anybody should be sold a rebuild. Legal AI adoption in small firms right now is a confidence problem before it is a capability problem, and confidence gets built by one small thing that works, not by a roadmap. My co-founder put the economics of it better than I do:
For lawyers, time is money and money is time. The big firms have all of these automations set up. It is the smaller firms that do not want to pay for another service and do not want to do this, and they just need a place they can go to save time and save money.
Read the middle clause twice. Do not want to pay for another service. A solo practice or a three-lawyer shop is already carrying a stack of monthly line items that each do about a ninth of a job. Another subscription is not a win. No more subscriptions stacking up is a win. Which is why the first thing you buy should replace something you are already paying for, not sit beside it. Work with what you already have, then replace it piece by piece, and keep the receipts so you can see which line items died.
Verticals beat platforms because attorneys will not configure
The sharpest market claim I have heard about legal software is this. The problem with a general-purpose tool is that lawyers do not love it, because they want it to be specific. A tool that could do anything requires somebody to decide what it will do. That somebody is the attorney. The attorney will not.
So the product ships as a practice-area edition, not a configurable platform. A small tool aimed precisely at one practice area’s workflow beats a large tool that could be aimed at anything. A general customizable edition can exist and ours does. It is the lower-value product, and pretending otherwise is how firms end up with a license, an empty configuration screen, and a renewal notice.
This is the same argument as the modular one, one level up. Narrow beats broad at the product level for exactly the reason narrow beats broad at the purchase level. Somebody has already done the deciding.
Configuration is unbillable, which is why it never happens
The reason attorneys will not configure is not laziness. It is economics. Configuration hours are unbillable, invisible to the client, and invisible to the partner reviewing the month. Setup time competes with billable time every single week and loses every single week. It is the same reason the learning never happens either, which is the whole argument in you cannot learn this while putting out fires.
Which is the argument for shipping something already loaded. The caselaw structure, the court-approved forms, the procedural sequence, the template language for the practice area, in the box. Not built by technologists who have never set foot in a courtroom. Built inside an active Florida family law firm and tested on real cases, real clients, real deadlines. The configuration is already done, because I did it on my own matters first.
If you want the map of what one thing to fix looks like at each stage, the five levels of AI in a law firm is the sequence I use on calls. And if your firm is already past the threshold, paying a stack of services to hold pieces of your own work product, that is the renting your own firm’s knowledge problem, and it is the signal to stop buying components.
Start with the catalog and pick one thing. It is organized by the job it does, not by tier, and the descriptions say plainly what each one will and will not handle. Browse the Florida family law catalog, and if you count six things you want, say so and I will tell you not to buy them separately.