“We’re able to vigorously represent the client.”

Brian was not new to AI. He drafted with ChatGPT for hours a day and had his own method: bring the cases he trusted, build a brief one section at a time, and carefully review the output for hallucinations and for the tells that make a filing look AI-written, like excessive bold text.

The drafting was fast. What came after was not. Every section still ended up pasted into Word, where the captions, index numbers, headings, signature blocks, and tables of authorities were rebuilt by hand, often late at night before a deadline. Preferences he had explained to the AI one day had to be explained again the next.

Describing that workflow in March 2026, before any of this changed, he put a number on it. “I spend 50% reformatting,” he said.

With DocketDrafter connected to Claude, every document comes back in Brian's format from the first draft: his font, his caption layout, and the right citation style for the court. Memoranda of law arrive with the table of contents and table of authorities already built, and he no longer checks the caption and signature block for errors that crept in.

The documents that leave his desk no longer vary. “All the letters that I'm giving to other attorneys to review are now just uniform,” he says.

He also reviews each document as it will look when filed, which is how he has always preferred to edit.

Brian did not build Projects, write skills, or learn a new settings screen. We connected DocketDrafter to his Claude account on one screen-share, then set up his format from a sample document and public filings from the firms he works for.

As he used it, he pointed out the details a careful litigator notices, and each one became a permanent default instead of another instruction to repeat: Tanbook citations in state court and Bluebook in federal court, a comma after the year in every mid-sentence date, his caption layout down to where the index number sits, and bold top-level entries in the table of contents.

We also connected CourtListener to his Claude, so he can research cases in the same session where he drafts.

The whole setup took one call. “It was just a matter of a Zoom meeting, you guided me, we shared the screen, I connected it,” he says, “and after that it was relatively easy.”

The time Brian saves goes into more work. The lawyers he works for now take matters that were not worth the hours before. One example is the motion to strike affirmative defenses under the CPLR, which he once considered not worth the squeeze. He gives Claude the complaint and the answer and asks for the full motion.

Back comes the whole package: a notice of motion, an attorney affirmation, a memorandum of law, and a proposed order, each one court-formatted for review. He then checks every case and every quote before sending the package to the attorney who will file it.

“Now you have a very fulsome motion to strike affirmative defenses,” he says.

Brian used to plan backwards from a filing deadline. Finishing the argument did not finish the memorandum of law: the table of authorities still had to be built by hand, several hours he had to find before the court's date. Now the tables build themselves and every document comes back in his format, so the last hours before a deadline go to the argument.

The same holds when something breaks late. If discovery falls through or a motion lands the night before a hearing, he can still get a polished letter to the attorney that evening, without waiting on anyone.