Connections: What AI Still Can't Do in Legal Writing

Where Great Legal Writing Really Begins

Writing for law firms shouldn't be about…well, legal minutiae.

It can be about anything, as long as it conveys an idea that relates to the firm.

This is the Jurassic Park legal writing argument:

If you needed a rudimentary understanding of DNA sequencing, would you rather wallow through a science textbook loaded with knowledge you are neither looking for nor can use, or would you rather watch Jurassic Park, be entertained, get a little educated as a side effect, enough to know you need a scientist going forward?

It's all about connections.

As in: making them.

I certainly didn't learn how to make them in some stuffy classroom or godawful book.

I doubt I would have learned a thing that way for the simple reason I would have been aware I was learning something, and, like Bart Simpson, I hate that.

Instead, I learned from the best, the British historian James Burke.

The Burke Chain: From Norman Stirrups to IBM

He had a show, Connections, on the BBC in the late seventies that found its way to PBS about the time Mike Liut shut out the Montreal Canadiens 1-0 in the playoffs, the generally acknowledged high-water mark of the franchise (see: Negative Whale).

He never taught.

He showed.

And he showed through connections.

The episode I've never forgotten started with a discussion of the Battle of Hastings in 1066 and the technological advantage the Normans had: the stirrup.

Ten or so connections later—some of the stuff seemingly out of the blue—and we had the computer.

Note: he didn't start with "here's how we developed the computer."

The viewer had no idea where he was going from "stirrups."

Now ask any AI you like to take you from the stirrup to the computer and guess what? It'll do it. It'll do it because it's read Burke's books and every webpage, blog post, and Reddit thread where he's mentioned. The chain is there.

"The AI isn't connecting the stirrup to IBM. It's remembering that Burke connected them. Not the same thing — and the difference is what counts."

It Can't Build the Next Chain

It can't build the next chain.

That's strictly a human-gotta-do-it thing.

It can't even reverse-engineer a connection piece unless it's [very] patiently taken through every step.

A while back, we wrote a client piece on the FTC's proposed ban on non-competes.

We didn't open with labor law.

We opened with the Titanic—too few lifeboats, the hearings, and the new rules putting more lifeboats on every ship.

Then the SS Eastland, in 1915, was tied to a dock in the Chicago River, top-heavy with newly required lifeboats.

It rolled over at the pier, and 844 people drowned ten feet from LaSalle St.

A congressional regulation was passed despite warnings by marine architects that while it was fine for ocean-going vessels, it would destabilize lake and river vessels.

The connection to a complete ban on non-compete agreements regardless of circumstances was a layup.

Months ago, I ran this through Gemini, Grok, Chat, and Claude.

Two hours of prompting and coaching and dropping hints a half-step short of saying it outright.

Two hours.

Google finally got it…

It cited our client's blog.

How to Stay Ahead

Want to be sure you're taken as human?

Make connections that AI can only find months after you post.

Which is, really, our entire job.

Don't go linear and don't ask AI for ideas—it'll suggest what everyone else is doing.

Turn the Terry Rozier NBA betting scandal into a franchise-contracts piece.

Find the prestige drama that is really about contract doctrine.

Find the treasure shipwreck that will explain courtroom procedure.

You can't prompt your way there.

You get there by reading, listening, or watching and…

"Oh, wait, that's really about that thing I do when I'm not reading, listening, or watching."

The Next Connection Hasn't Been Made Yet

James Burke didn't start with the computer. He started with a stirrup and trusted the audience to stay with him until the pieces fit together.

That's how we approach legal writing.

Not by asking what's already been written, but by asking what no one has connected yet.

If your firm's content has started to sound a little too familiar—or a little too much like everyone else's—schedule a conversation. We'd rather build the next chain than repeat the last one.