On a recent consult, an owner shared her screen and pointed at a lead her new AI had marked “Chase Hard.” “Look — it even tells me who’s worth my time,” she said. Then she scrolled. The lead was three weeks old. Nobody had called. The AI had done its job perfectly: it found a great case, scored it, flagged it. And then it sat, because nothing behind the score was built to move.
She’d turned on the smartest part of her system and bolted it onto the weakest.
I get some version of this on almost every consultation now. Lawmatics has been rolling out real AI — Merlin Qualify (originally launched as QualifyAI) scores your leads, and newer tools in the same suite go further and start contacting them. If you’re on Lawmatics you’ve seen the announcements, or a colleague has asked whether you’re “using the AI yet.” So let me answer the question directly, and it’s probably not what you’ll hear from whoever sells you the software.
Whether you should turn it on depends almost entirely on what you’ve already built underneath.
What Lawmatics actually launched
To their credit, this isn’t a gimmick. Lawmatics was public about not shipping a surface-level “we wired in ChatGPT and called it AI” feature, and they built it alongside nearly a hundred firms over several months. (Lawmatics’ own announcement of the Merlin AI suite lays out the pieces.) Understanding what it is tells you how to judge it.
In plain terms, the lead-scoring piece reads your firm’s case history, learns which patterns made past cases worth taking, and gives each new lead a recommendation you can act on — categories like Chase Hard, Chase, Refer, Reject — instead of a bare number. Each one comes with its reasoning and a confidence level, so it isn’t a black box. It trains on your history and your criteria, it can ask a lead for missing information, and the score can trigger your workflows automatically — route a referral, or fire off an immediate-contact sequence for a high-priority case.
Here’s the detail that tells the whole story. When it first launched in October 2025, Lawmatics deliberately left out an AI that would call leads by phone — because firms told them their clients weren’t ready to have a robot as their first experience. Nine months later, in July 2026, a new tool in the same suite — Merlin Engage — added exactly that: agentic AI that engages leads by text, email, website chat, and phone (currently in beta). The technology was ready before the client relationship was, and they waited until the readiness caught up. That patience — letting the process lead the technology — is the entire point of this article.
The one idea that ties it together
If I had to reduce this to a sentence: AI amplifies your process. It doesn’t repair it.
Lead scoring rests on top of the intake you already have. It reads the data your intake collects, applies the criteria your firm has defined, and triggers the workflows your firm has built. Those are all already yours. If they’re clear, AI makes them faster and more consistent. If they’re vague, AI won’t tidy them up — it just runs them at higher speed. It’s the pattern I see in every audit: the tool is almost never the problem; the process feeding it is. AI only makes that truth matter more, because now the output arrives with a score that looks authoritative.
That’s the real risk. A broken intake at least looks broken. A broken intake with an AI score on top looks trustworthy — right up until you notice the “Chase Hard” leads are aging out untouched.
When it genuinely helps — and when it doesn’t
AI lead scoring pays off once a few things are true. When your lead volume is high enough that deciding who to handle first is a real cost. When you already know what a good case looks like and it’s written down, not just living in your head. When your history is organized enough to learn from. And when you already have workflows built for the score to trigger. With those in place, turning it on usually pays for itself.
It’s better to wait when the opposite is true. When “a good case” was never actually defined, the AI guesses at your judgment and you start trusting the guess. When your history is messy — inconsistent tags, empty fields — it learns from that mess and hands you a confident answer resting on unreliable data. When you flag a lead as priority but there’s no follow-up behind it, the score is just a more expensive way to lose the same lead. And if you’re hoping AI replaces an intake team you haven’t trained yet, slow down — it reprioritizes the work, it doesn’t replace the human conversation. Lawmatics said that themselves.
What has to be in place before you turn it on
I think of it as three layers, from the bottom up. First: clean data and clear criteria — tagged sources, consistent fields, a written definition of a qualified lead. Second: stages and automations that actually work. Only then, third: the AI on top, with good data to read, real criteria to apply, and real workflows to trigger. Most firms want to start at layer three because it’s the exciting one. The firms that get value start at layer one.
Before you pay for the add-on, four questions. Can you write, in one paragraph, what makes a lead worth pursuing at your firm? Is your last twelve months of cases tagged consistently enough to learn from? When a lead is flagged priority, does something automatic already happen — or does it depend on someone remembering? And are you adding AI to sharpen a system that works, or to cover for one that doesn’t? Answer the first three clearly and the AI will likely pay for itself. If you can’t, the highest-return move isn’t the add-on — it’s getting those answers in order first.
Because “AI-native” doesn’t mean the AI does the work. It means the process, the data, and the automations are built so AI can sit on top and amplify them safely. That underneath work — defining criteria, cleaning data, building the workflows the score will trigger — is the least glamorous part, and it’s exactly what makes the shiny layer actually pay off.
If you’re weighing whether to turn on Lawmatics’ AI, the honest first step isn’t the add-on — it’s checking whether those three layers are solid underneath. That’s where our Customized AI Solutions work starts: clean data, clear criteria, automations that hold, so the AI you switch on is amplifying a system that already works.
So before you turn it on, ask yourself the real question: are you adding AI to a system you can already trust — or hoping it becomes the system you never built?
