Why We Chose to Build a Law Firm, Not a Legal Tech Vendor
I've been asked by numerous friends and investors over the last couple months why we chose to build a law firm and not another legal technology vendor. It's a fair question. We come from Casetext which built legal AI products for lawyers for a decade before selling to Thomson Reuters, and all three of our founders are engineers.
I think the answer becomes self evident, though, if you look at Anthropic's latest release.
Last Friday evening, Anthropic dropped a Microsoft Word plugin geared toward lawyers. I took a look, and as with most things Claude, the UX is awesome. Specifically, they seem to have embedded their web app’s React components into Office's webview container, so that the Word plugin feels exactly like the web app. They also built a few simple agentic tools that could be called from the chat: one that found minor errors with high precision and recall in my test called “check-doc," and one that totally screwed up by misreading section numbers and paragraph spacing called “copy-edit."
I reported back to my team that the plugin was as-of-yet unusable, though, because the redlines were made in blocks at the sentence level and not at the word and character level, a dealbreaker for lawyers redlining documents because a single changed word or number could make a big difference. Furthermore, there was no playbook feature to provide structured client background to the AI. A firm's own playbook could normally provide a set of predefined positions, fallback language, and client-specific preferences guiding how a tool drafts and flags issues.
Fixed in 72 Hours
Remarkably, the next day, only 72 hours after the plugin was released on Friday night, the redline issue had been entirely fixed by Anthropic. I also discovered that the plugin’s “skill creation” feature had always allowed users to build custom playbook skills. The tool went from unusable to the the best Word plugin I have seen on the market in one day, and our firm may well integrate it into our workflow soon.
Building for the Outcome, Not the Tool
This brings me back to why we decided to build a law firm instead of a legal technology company for lawyers. If we did the latter, Anthropic diving headfirst into making lawyers more efficient would undermine our business. We would be carefully designing our roadmap to entrench tools that Anthropic can’t or doesn’t want to build, while scouring for information on what’s coming next. How lame.
Instead, tools like this just make our services better, faster and cheaper, because clients are our obsession, not lawyers. Tools like this one will ultimately drive our price point down into the 92% of unmet legal demand in the US, bringing us more clients and creating immense value for people and businesses.
That is our mission and we are entirely agnostic about which tools we use to achieve it.
TL:DR: General Legal chose to build a law firm instead of another legal technology vendor because when the company behind the underlying AI model also builds tools for lawyers, a vendor's roadmap can get overrun overnight. That risk became concrete when Anthropic released Claude for Word in beta on April 11, 2026, aimed squarely at lawyers, and fixed its most significant early flaw within about 72 hours. A law firm has nothing to lose from a better Claude. General Legal builds around client outcomes rather than defensible software, using whatever tools work best, to chase the 92% of civil legal problems that low-income Americans don't currently get adequate legal help for, according to the Legal Services Corporation's 2022 Justice Gap study.
FAQs
Why did General Legal build a law firm instead of a legal technology company?
Because a legal tech vendor competes directly with whatever the underlying AI labs, like Anthropic, decide to build for lawyers next. A law firm has no such conflict: any improvement to the underlying AI just makes the firm's own service better, faster, and cheaper.
What is Claude for Word, and why does it matter to this argument?
Claude for Word is Anthropic's Microsoft Word plugin, released in public beta on April 11, 2026, with legal contract review as its headline use case. Its rapid launch and near-daily improvement is the concrete example General Legal points to for why building tools for lawyers is a risky bet when the model maker itself might build the same tool.
What was wrong with Claude for Word when it first launched?
Its redlines landed in blocks at the sentence level rather than the word and character level, which doesn't work for how lawyers actually redline documents, and there was no obvious way to feed it structured client background through a playbook.
How fast did Anthropic fix the redlining issue?
About 72 hours, going from Friday night's release to a fix the day after the firm's testing.
Where does General Legal's founding team come from?
All three founders of General Legal previously worked at Casetext, a legal AI company that built products for lawyers for a decade before Thomson Reuters acquired it for $650 million in 2023.
What does General Legal mean by "92% of unmet legal demand"?
This statistic is drawn from the Legal Services Corporation's 2022 Justice Gap study, which found that 92% of the civil legal problems substantially affecting low-income Americans did not receive any or enough legal help. General Legal's ability to offer lower prices, driven by better AI tools, provides a way to reach more of that unmet demand.