The next billion dollar law firm will be built by someone who's never practised law. Their best lawyers will run on a local machine. One £250k compliance lawyer will oversee everything. They won't even know about the wastage and bloat the rest of the industry treats as normal.

Same lesson in legal services. The founders worried about Anthropic eating their legal startup are arguing about the wrong failure mode. The firms that fail will fail because they never changed the operating model, not because a lab ran them over.

Twenty points of annualized growth between the AI heavy and the AI absent is not a gap. It is a restructuring. Legal services is about to show the same fault line. The firms that run agents in production are moving matters through the pipeline at a rate the seat licensed copilot shops cannot model, let alone match. Three years of this and the league table looks completely different.

The friendliest place on earth. Ready for Morgan Wallen at Bryant Denny Stadium tomorrow 🤠 #RollTide

Was talking about Ai pain points today with a friend and the black hole you can end up in when brainstorming new ideas with Ai. Anyone else had the wind taken out of their sails by Ai?

I’ve got a bipolar amnesiac personal assistant. The biggest pain with integrating Ai into your business life is amnesia. You spend hours planning the idea, all the steps to build the idea, the research, the accounts you need to create etc. Go to sleep. Compaction happens. I might as well have been talking to myself all day.

Anthropic must be about to drop a new model as Opus 4.6 has noticeably deteriorated in quality output over the last 2 weeks 💩

52% of corporate legal departments now have AI tools. Only 7% report lower costs. That gap tells you everything. The tools work. The operating model doesn't change. The companies that will win in legal AI aren't selling copilots to law firms. They're owning the outcome. Taking the case from intake to resolution. The data, the operations, the regulatory license, the result. Buy the outcome, not the tool. Exactly right.

The hardest problem in production agents isn't intelligence. It's memory. Every agent wakes up fresh. No context from yesterday. No lessons from last week. Your best agent is an amnesiac. We solved this with markdown files per agent per day. Now OpenClaw is building it into the platform with light, deep, and REM style consolidation. This is what separates toy demos from agents that actually get better over time.

Today's Anthropic news is the best argument for multi provider infrastructure I've seen. We run legal agents in production. If any single vendor disappeared tomorrow, our cases keep moving. API keys, local models, open source tooling. The companies treating vendor access like a utility are about to learn it's actually a dependency.

Harvey building an internal "law firm world model" is exactly where this is heading. The firms that win won't be the ones buying copilots. They'll be the ones whose agents already know their workflows, their data, their edge cases. That's not a software purchase. That's an operating model.

Everyone talks about AI replacing jobs. Nobody talks about AI creating markets that didn't exist. Access to justice is a huge issue. There are tens of millions of people globally with valid legal claims who will never see a lawyer. Rural communities. Low value cases. Populations that big firms can't serve because the overhead doesn't justify the fee. When AI drives the cost to service a client close to zero, those people become viable clients for the first time ever. That's an entirely new market. Massive in scale. Completely untouched. Access to justice is no longer a charity play. It's the biggest greenfield opportunity in legal with a great outcome for all.

This is what operating on AI actually looks like. Not a chatbot. Not a copilot. An entire operating model rebuilt around agents. We run a regulated law firm the same way. Specialist agents handle every stage from intake to compliance. The org chart looks nothing like a traditional firm. It looks like a tech company. The gap between "experimenting with AI" and "running on AI" is where 99% of companies are stuck right now.

Good weekly roundup. Crosby pitching Big Law on R&D is interesting. The firms that treat AI as an R&D investment rather than a cost centre will be the ones still standing in five years. Legora hitting $100M ARR in 18 months tells you everything about demand. The market isn't waiting for anyone to figure it out.

Three lanes are forming in legal AI right now. 1. Copilot platforms selling to law firms. Harvey and Legora are winning this. 2. AI native firms buying practices for the data. Multiple companies independently converging on this model. 3. Incumbents building their own models. Thomson Reuters just announced theirs. The interesting question is which lane creates the most value. The copilots are selling picks and shovels. The AI native firms own the mine.

$60bn is the autopilot opportunity. But Sequoia is looking at where AI replaces existing work. The number nobody is modelling is the greenfield. Tens of millions of people with valid claims who never had access to legal services because the economics didn't work. That market didn't exist before AI. It's about to.

Sequoia nailed it. "Services is the new software." We're living this in legal. AI agents now handle case triage, document processing, compliance checks. Work that used to require teams of paralegals. But the bigger opportunity isn't replacing existing services. It's serving markets that were never economically viable before. Tens of millions of people with valid legal claims who've never spoken to a lawyer. AI changes that math completely.

OpenClaw going enterprise through Microsoft 365 is massive. We already run specialist agents across our entire legal operation on OpenClaw. Case triage, compliance checks, data processing, client communications. The fact that this is now getting a Teams integration means every knowledge worker could have what we built for legal. Personal agents that actually do the work. Not chatbots. Not copilots. Agents.