Writing
The long-form thinking behind the work, by Dominic 'Dom' Conte - on what makes legal innovation stick, and what quietly makes it fail. A new piece roughly every fortnight.
Building for the profession you left gives you a head start that's hard to match. It also gives you blind spots, a network that cuts both ways, and an identity problem nobody warns you about.
ReadMost law-firm AI policies are written for the risk committee and read by nobody else. A practical guide to writing one that works at 6pm, when a fee earner is deciding whether to paste something into a tool.
ReadAI is taking away the repetitive work that used to turn trainees into good lawyers. The pipeline won't fix itself, and firms that don't redesign how juniors learn will discover the gap years later, when it's too late to close.
ReadModels are becoming a commodity, and every serious legal tech company knows its real defensibility lives in data. Almost nobody says so out loud, because the honest version raises an awkward question about whose data it is.
ReadMost firms are treating AI and client confidentiality as an internal compliance problem to be solved quietly. It isn't. It's a conversation with the client, and the firms that start it first will earn the trust everyone else is waiting to be given.
ReadI don't regret leaving practice, and I wouldn't go back. But there are things about being a lawyer I still miss - and paying attention to them has made me better at building for the people who stayed.
ReadAlmost every legal AI business case rests on time saved, and almost nobody measures it properly. A practical framework for working out whether a tool is genuinely making your lawyers faster - or just moving the work somewhere else.
ReadLaw firms have every reason to move slowly on AI, because the billable hour rewards them for it. Their clients' in-house legal teams have the opposite incentive - and they are the ones who will make the change happen.
ReadPartners sign the contracts, but trainees and junior associates do the work. Legal tech built around what partners ask for tends to go unused - the products that stick are designed for the person actually at the keyboard.
ReadAI document review is genuinely good at the documents that look like every other document. The risk lives in the ones that don't - and that long tail is where legal AI is won or lost.
ReadThe first time I had to put a price on software after leaving practice, I reached for the instincts of a fee proposal. Almost every one of them was wrong - here's what I got wrong, and what I'd do now.
ReadMost legal AI projects fail before a line of code is written, because nobody scoped the use case properly. Here is the one-page framework I use to decide whether an idea is worth building.
ReadInnovation teams treat the sceptical partner as an obstacle to manage. In my experience, the partner who says no is usually carrying information the project needs - you just have to decode it.
ReadWeek one of a legal tech rollout almost always goes well. Week two is where adoption is actually won or lost - and almost nobody plans for it.
ReadMost people judge an AI first draft by how finished it looks. That's the wrong test. A good first draft is the one that's fastest and safest to review - and it has a recognisable shape.
ReadLegal training gives you real advantages as a founder - and installs instincts that will quietly kill a product company. These are the habits I had to consciously unlearn, and the ones I kept.
ReadEvery legal AI demo is designed to impress you. Almost none are designed to inform you. A practical guide to the questions that separate a genuinely useful tool from beautifully rehearsed theatre.
ReadThe legal AI everyone talks about has a chat box and a brand. The legal AI that actually wins will have neither - it'll disappear into the tools lawyers already use, and that's exactly why it'll work.
ReadAs AI does more of the legal work, the value concentrates into a thinner and thinner layer of genuine judgement. That thin layer isn't a problem for the model to solve. It's the entire business.
ReadMost legal AI pilots die quietly between the demo and the rollout. The reason is rarely the model. Here is what separates the pilots that scale from the ones that don't.
ReadShould your firm build its own legal AI or buy off the shelf? A practical framework for general counsel and managing partners, from someone who has done both.
ReadSelling software to lawyers is its own discipline, and most of what works when selling to other markets actively backfires here. The lessons I learned the hard way, building and scaling in the most sceptical buyer market there is.
ReadNot the science-fiction version with robot lawyers, and not the cynical version where nothing changes. A concrete, specific picture of how legal work will actually be delivered five years from now.
ReadThe innovation team has become the way law firms convince themselves they're changing without changing. It's a comfortable substitute for the real thing - and the real thing is just shipping.
ReadEvery legal-tech case study is a success story. The honest ones - the tools that got switched off - teach more than the wins. Here's one, and what it taught me about why good tools die.
ReadMost legal software is technically fine and practically dead - bought, deployed, and never opened. A field guide to the unglamorous design decisions that separate the tools fee-earners reach for from the ones they ignore.
ReadEveryone wants to talk about the lawyers and the AI. The people who'll actually decide which firms win the next decade sit in legal operations - and almost nobody is paying them enough attention.
ReadEveryone's waiting for the AI-native law firm to emerge from inside the profession. It won't. The thing that eats a category usually doesn't look like the category - and the legal version won't look like a firm at all.
ReadPredictions are cheap, so here are six specific, falsifiable ones for legal AI this year - the kind you can hold me to in December - plus one I'd genuinely rather not come true.
ReadI spent a week sitting next to lawyers watching them actually work, not telling me about their work. Almost everything useful I now believe about legal software came from the gap between the two.
ReadThe relationship between software and professional services isn't a clean replacement. It's a long, quiet erosion followed by a sudden collapse - and law is somewhere in the quiet part right now.
ReadEveryone watches the top of the firm for signs of change. The real shift in legal economics is happening at the bottom - in the work juniors do - and it quietly breaks the model the whole firm is built on.
ReadThe legal AI industry obsesses over accuracy benchmarks. But lawyers don't adopt tools because they're accurate. They adopt tools they can trust - and those are two very different properties.
ReadAgents are the hype of the moment, and most of the legal demos are fantasy. But there's a real, narrow place where agentic workflows arrive first in a law firm - and it isn't where the conference talks point.
ReadMost firms aren't wrong about AI because they're behind. They're wrong in three specific, expensive ways that even the keen, well-funded ones keep repeating.
ReadFirms keep trying to solve adoption by buying better technology. The technology was never the bottleneck. The bottleneck is getting a good tool into a busy fee-earner's actual workflow - and that's a distribution problem, not a tech one.
ReadMost law firms think they're building products. They're running projects with a product's marketing. The difference isn't pedantic - it decides whether the thing survives contact with the second client.
ReadVertical AI is the phrase every investor and vendor now uses about legal. Most of them mean a model with a legal logo on it. In a profession that sells judgement, the vertical is somewhere else entirely.
ReadEveryone predicts the death of the billable hour and keeps being wrong. The hour isn't dying. The work it used to price is being pulled apart, and the hour only survives on one of the pieces.
ReadI didn't leave practice because I disliked law. I left because I kept watching good lawyers do work that a piece of software should have done years ago - and nobody was building the software.
ReadEveryone is waiting for the moment law becomes a product. It already has, in a hundred small places at the edges of the firm. The shift is just badly distributed.
Read