For thirty years the legal profession has chased something it could never quite name or build: the productized service. AI is making it unavoidable, for the firms and for me. Yes, I am doing it to my own practice, in the open, and this is the field report.

I usually send you an argument. Today I am sending you a field report, including the parts that did not go my way. Give it ten minutes.

TLDR: I am taking the advisory work I have done by hand for twenty years, decoding how a partner actually wins and turning what is moving in their market into their next move, and productizing it into a system any partner can use. Then I put it in front of real partners and tracked the only thing that counts, not whether they liked it, but whether they did anything differently. The result is honestly mixed, which is what a real test looks like. And the most useful idea in here, worth more than the rest of the issue if you are weighing any AI for your own practice, is the line that separates real productization from theater: measure behavior, not applause.

THE FULL BRIEF

What a Productized Service Actually Is

Let me start with the term, because the profession throws it around loosely and almost no one has actually built one.

A productized service is what you get when you take work that used to live entirely inside an expert's head and an expert's hours, and turn it into something repeatable, packaged, and deliverable without that expert in the room every time. Same quality. Defined scope. Consistent output. It stops being "hire me and we will see" and becomes "here is the thing, here is what it does, here is what it costs." Think a hybrid: part product part service.

Firms have flirted with this for decades. I know I used to get hired to help them. Fixed-fee menus, legal ops, document automation, the alternative providers. Mostly it stalled, for two reasons. They never really understood what a productized service was. Was it technology? Templates? Dashboards? Nope. Regardless, underneath these attempts was always a quiet threat, because productizing a service means decoupling the value from the individual who delivers it, and the individual partner is exactly what a firm's economics and status are built on. You cannot productize the indispensable rainmaker without admitting the rainmaker is, in part, productizable.

That used to hurt. How firms are rushing toward it because AI is removing the excuse. The hard part, encoding expert judgment into a system that reproduces it, has collapsed in cost. Now the question is no longer whether legal services get productized. Harvey and the rest are already doing it to the work product, which was the subject of my last edition on Deep Coverage and of Issue #279. The only question left is who productizes deliberately, on their own terms, and who has it done to them.

I am not immune. My advisory practice is exactly the kind of bespoke, judgment-heavy, "you could never package this" service everyone assumes is safe. It is not. Rather than wait, I am productizing it myself, on purpose, to get ahead of the curve instead of behind it. And I am going to show you what that actually looks like, including the parts that did not go my way, because that is how we all learn.

What I Am Building, In Plain Terms

Before the guts and the feedback, here is the concept in one breath, because some of you are new and the rest deserve the refresher.

I am building the thing I have done by hand for twenty years, as a system. Take a partner. Decode how they actually win: the clients they are built for, the work they are genuinely best at, the handful of market signals that tend to precede their best matters. I call that their practice genome. Then point it outward, every week, so that what is moving in their market gets weighed against who they are and handed back as a specific move, not something that overwhelms or get ignored. Something they immediately take a bite of and swallow.

It has two parts. The front door is a structured interview that decodes the genome and hands the partner a portrait of their own practice that most have never seen. The engine is a weekly desk that watches their market and turns the noise into a few moves, each ending in a single decision. The interview teaches the system who the partner is. The desk uses that to tell them what to do this week. Together they are the coverage desk I wrote about last time, creating and nurturing the  senior partner's instinct for being ahead of a client, rebuilt so it no longer depends on that partner having infinite hours to stay current.

That is the whole idea. Everything below is how I am building it, what partners did when they touched it, and where it broke.

The Discipline Most AI Skips

You know this but sometimes you get dazzled and forget. The demo is not the product. The applause in the room is not the result. A partner saying "wow, that's impressive" tells you nothing about whether the thing changed how they work, and changing how they work is the entire point.

For my own project, when I started building and testing, I made one rule. I would not measure reactions. I would measure behavior. Did the partner reread it. Did they keep going and push on it. Did they actually do something differently. Did they tell someone else, without me asking. Applause is free and polite. Behavior costs something, and that is what makes it evidence.

That rule is the most useful thing I can hand you in this issue, whether or not you ever touch anything I build. The next time someone shows you a legal AI tool, do not ask yourself whether it is impressive. Ask whether anyone changed their behavior after using it. Most of the answers go quiet right there.

What I Actually Built

Let me lift the hood, because the way it is built is part of the argument, and because this is what productizing my own judgment actually took.

It is not one chatbot answering questions. That is the version everyone demos and almost no one ships into real use. What I built is a set of agents, each with one job. One reads the structure of a partner's practice across a defined set of dimensions and scores it. Another builds a dossier from the public record while the partner is still answering, so the system is not relying on self-report alone. A third carries the conversation afterward, when the partner wants to push back and go deeper.

And sitting across all of it is a layer I still orchestrate by hand for now. I say this plainly because it matters: the judgment about which signal is worth a partner's attention cannot be automated yet, and pretending otherwise is how you get a confident tool that is confidently wrong. Building has never been easier. Knowing what is worth building, and which signal actually matters in a specific practice, is now the scarce thing. That is the same taste a great rainmaker has, and it is the part I refuse to hand to a model right now.

***I say "I" but I should say "we." I have a silent partner in this. One of us owns the methodology and the judgment. The other owns the systems that run it. Neither half works without the other.

How I Tested It

Before anything reaches a partner, it runs against an evaluation framework I built. The scoring is not "does this sound good." It is "is this accurate, specific, and useful." Sounding good is the trap. A fluent paragraph that says nothing true about a particular practice is worse than useless, because it spends trust I do not get back.

Then I calibrate it partner by partner. Not against a generic template of "a litigator" or "a deals partner," but against the actual person in front of me. This is the part I would tell any partner to demand of anything they are sold. Generic input produces generic output, and generic output is exactly what has trained partners to ignore their firm's so-called intelligence tools for years.

Then I track what happens after. A decline to act or reconnect with me is not a failure in my book. It is a signal that the read was wrong, and the system adjusts. That is the loop. Build, watch real behavior, correct, repeat.

What the Partners Did

Here is the honest field report.

Almost no one just read their profile and stopped. They went into the deep chat conversation behind it, added context the interview could not capture, and actually argued with it in some cases. One partner told me the profile had missed what actually mattered about his practice, but once he could talk to it and add what was in his head, the read changed substantially. That is a finding about the limits of any intake form, and I will come back to it.

In the one-on-one sessions I conducted to hear their reactions and feedback, I learned their behavior went further than I expected. One partner rewrote his own positioning off the back of his profile telling him he needed to, drafted new material himself, and sent it to me asking for edits. Another told a colleague to take the interview, with no prompting from me, and the colleague signed up. Partners do not do homework for a tool, and they do not refer a tool to a peer, unless it showed them something real. That is behavior, not applause, and it is the strongest signal I have.

The written reactions ran the full range, which is exactly what you want from a real test and never get from a staged one. "Surprisingly accurate." "A strategic mirror, direct and useful." And also, from another partner, "it got the broad strokes but missed what matters." One told me it beat the human BD coach assessments he had done before because it did not just drop him into a predefined bucket or archetype, which is what most of these tools quietly do.

And a couple of partners told me, bluntly, that the profile read too much like AI. All "it is not this, it is that" with em dashes. They were right. That is getting fixed. I am telling you this because a field report that only contains the flattering parts is not a field report. It is a brochure.

What Failed, and What It Taught Me

The deepest lesson was not a bug. It was structural. On its own, the tool is a mirror. A sharp one. It shows a partner their own practice with a clarity they rarely get, and that turns out to be genuinely useful for getting focused. What a mirror cannot do is show you the room behind you. It cannot deliver the jolt of seeing something about your market you could not have known. That requires the mirror plus a window, the partner's own data colliding with what the market is actually doing.

And more than one partner named the same gap, almost word for word. The insight is fine. The problem is execution. "You identified the symptom, but there was no plan to fix it." "It is good for ideas, but for those of us who already have ideas, the trouble is execution, and that takes time. Solve that and it is valuable." You got it!

Read that again, because it is the most important thing in this issue. They asked for someone to do the work of doing. That is not a complaint about the tool. That is the market telling me what to build next, in its own words. It is the difference between a profile and Deep Coverage, between knowing the move and making the move, and it is exactly the layer I am building now.

There were ordinary failures too. The conversation promised one partner it would update his profile and then did not. Betas break, and the breakage is data. But the structural lesson is I have almost nailed the profile generation because partners want to keep going to take action after they read it. They are not ignoring it.

Well, I Built the Next Piece, and Tested That Too

The execution gap is not a someday problem. It is what I am building now, and I put the first version in front of a partner this week.

It is a weekly analyst page. One page. Five signals from the partner's own market, no more. Each one carries a freshness read, a clickable source so it can be verified in seconds, a plain why-now that answers the reflex "is this a hallucination," and then the part that matters: the specific move to make and the actual words to open it. Every item ends in one decision. Do it, skip it, or hold. It is the opposite of the firehose. It is one page that respects the only resource a partner cannot make more of.

I need to emphasize something here. This analyst page is designed to be consumed by only one person; that partner. It is personalized for them based on their genome. The language and tone. The insights. The choices to take action. Everything is personalized and designed for that one partner. Every partner is one of one in my eyes, and the agents’ eyes.

The first partner to see it, a practice leader with a serious book, wrote back in four sentences. "Wow. This is great. Just the type of stuff I am looking for. Would love to discuss refining it."

Now hold that against the rule I opened with. The "wow" is applause, and applause is free. The evidence is the last sentence. He wants to build it with me and put his own clients and targets behind it. That is behavior, and it is exactly what separates a clever demo from a system a partner runs their week on.

Why I Am Showing You This

Two reasons.

First, because the thesis is holding up in the only currency that counts. Partners are not politely nodding. They are rereading, rewriting, referring, and telling me what is missing. That is a profession that has been starved of real strategic support for a long time, reacting the way starved people react when you finally put something real in front of them. This connects to something I have believed since founding Bold Duck, whose name stands for Business of Law Designed. Most high-performing partners suffer from what I call Strategic Unconsciousness. Untapped advantage sitting right there, invisible to the person who has it.

Second, because the profession deserves AI held to a higher standard than the one it is being sold. Demos and vibes are not evidence. Behavior is. If everyone evaluating legal AI asked the simple question I built my whole test around, did anyone do anything differently afterward, the hype would get a lot quieter, and the real work would get a lot easier to find.

Here is the question I will leave you with, the same one I have been asking partners.

What would change in your practice if the few things you know matter, and never have time for, arrived on one page every week, each already turned into a move you could make that day? Not another tool to operate. Not another news feed or blah blah blah report. A personally designed system built around you, tested against your reality, and measured by one thing only: whether you acted.

This is my productized service and building it out is what comes next. I am doing that with a small number of partners willing to put their own world behind it, the way the one above wants to. If that is you, reply and tell me about your practice. I would rather build it with a few people who will actually use it than show it to a hundred who will not.

Either way, hear the larger point under all of this. The productizing of legal services is not a forecast. It is already happening, to the work product first and to everything else next, including the parts of this profession that have always told themselves they were too bespoke to package. The only real choice left is whether it happens to you or through you. I made mine. I am productizing the thing I do best before the wave does a worse version of it for me. That is the move I would not wait on.

Talk soon again, Josh

Over the last 18 months, MavSource’s founder tracked and invested around major AI-related names like Micron +100%, Nvidia +74%, Sandisk +130%, Western Digital +74%, TSM +22%, Broadcom +27%, Okta +35% and Lam Research +39% — an average return of approximately +63% across the group.

Now, he’s bringing the sources and ideas behind those insights to you in one daily email digest. MavSource aggregates the most important AI updates from newsletters, podcasts, company news, AI labs, funding rounds, and more — then summarizes what matters in a simple 5-minute brief.

Subscribe For Free Now

Past performance does not indicate future results. Informational only; not investment advice.

To read previous editions, click here.

DISCLAIMER: None of this is legal advice. This newsletter is strictly educational and is not legal advice or a solicitation to buy or sell any assets or to make any legal decisions. Please /be careful and do your own research.