AI makes bad lawyers worse

Wed, 02 Sep 2026 06:06:54 GMT

Daniel Lord-Doyle co-founded Mary Technology after watching his co-founder Rohan struggle through his first weeks as a lawyer, buried under documents with no time to actually help the client in front of him. Mary started as a glorified recorder for client meetings. It's now a fact-extraction engine that litigation teams use to work through hundreds of thousands of pages without missing the one line that matters.

We get into the "negative space problem" (how do you know what the AI missed), why adversarial review has to happen at every stage rather than just at the end, and a genuinely alarming Wharton study on cognitive surrender, people trusting a wrong AI answer 80% of the time, even when they're being paid to get it right. Daniel also makes the case for why he thinks the legal industry ends up with more lawyers, not fewer, once AI properly lands. Oh, and there's a solid ten minutes on Birmingham secretly being Mordor.

We dive into:

* Why lawyers are about to become more valuable

* How to prove AI reviewed everything and didn't ignore all the hard bits

* Adversarial review throughout the matter lifecycle

* People trusting AI's wrong answers even when $$$ is on the line

* Why general-purpose AI is roughly 100x more expensive and 100x lower quality than point solutions

* Why AI means we'll need more lawyers in five years

Chapters:

0:00 Cold open: fewer lawyers, cognitive surrender

0:51 Welcome and growing up in Birmingham

3:21 The cello years and Tolkien's Birmingham

7:15 Durham hustles and first sales jobs