Traditional and A.I.-Powered Brief Drafting Go Head-to-Head

Google+ Pinterest LinkedIn Tumblr

We wanted to put A.I. in law and legal practice to the test, so we set up a head-to-head challenge to find out if attorneys can write better legal briefs with AI. On one side, we had a products liability litigator (Danielle Brim) with over a dozen years of experience, who would show how to write a legal brief the traditional way. On the other side, we had an attorney (Casetext’s CEO, Jake Heller) who has never worked on a products liability case, but would be armed with Casetext’s A.I.-powered drafting technology, Compose. 

They were both given this prompt and asked to see how far they could get in 10 minutes of work:

You represent Greenwood Pharmaceuticals, a manufacturer of a prescription drug, Alvin, which the FDA approved for the treatment of migraine headaches in adults. You’re preparing to move for summary judgment in a case filed by a plaintiff, Jessica Wilson, in the Supreme Court of New York. She alleges she used the drug daily to treat her migraines for several years, and as a result, developed adverse cardiovascular side effects. She alleges that Greenwood failed to adequately warn about the risks and asserts claims for strict liability, failure to warn, negligent failure to warn, and a violation of the general business law sections, which are the consumer protection statutes. 

You now want to move for summary judgment based on three arguments: 

  1. That the warnings were adequate as a matter of law, because the labeling expressly warned about an increased risk of cardiovascular side effects
  2. Ms. Wilson’s prescribing physician testified that he still would have prescribed the drug even if they provided different warning because her migraines were really frequent and debilitating 
  3. Ms. Wilson’s claims under General Business Law fail as a matter of law 

The head-to-head challenge was moderated by Casetext’s General Counsel, Laura Safdie, who checked in on each participant’s progress every couple of minutes. (Watch the full head-to-head challenge below.)

Getting started

  • Compose: With Compose, Jake quickly got started by selecting the Motion for Summary Judgment for Pharmaceutical and Medical Device litigation from the Compose Motion Library, then entering some basic info like his jurisdiction and the names of the parties. After that, he was able to start scanning the list of available arguments. He reviewed the standard for summary judgment in New York and read over the Practice Note provided by the Compose editorial team. He then began building his argument section by selecting pre-packaged arguments along with summary judgment standards. 
  • Traditional drafting: Danielle started by identifying briefs from other cases in the same jurisdiction and with similar claims. In a real-world setting, she said, she would build up this folder of briefs by checking in with colleagues, searching her firm’s DMS, and potentially reaching out to co-counsel, and the process of gathering these briefs would likely take much longer than the two minutes allotted so far. 
Jake and Danielle show their progress a couple minutes into the drafting process.

Digging in

  • Compose: A few minutes in, Jake had built out his opening summary judgment statement, found the failure to warn section, and started to build out that argument section of his brief. He also discovered through a Compose Practice Note that strict liability claims and negligence claims are generally treated as equivalent in New York, found the appropriate four-part test, and added some legal standards to his draft. 
  • Traditional drafting: Danielle, meanwhile, had started digging into one of the related briefs. She found where the summary judgment standard was articulated in that brief and copied it into a Word document, leaving herself a note to later go back and check that the rule hasn’t changed since the brief was filed, and another note to make sure a cited case is still good law and potentially find a more recent citation.  

Final progress

  • Compose: In the last couple minutes, Jake started to tackle the question of whether Wilson can establish proximate causation. To support his argument here, Jake added this sentence to his draft, “The Doctor would have prescribed Alvin even if she had different warnings regarding its potential cardiovascular side effects,” which he was able to use as a Parallel Search query. Parallel Search returned a few cases to support Jake’s argument, and he selected one to add to his draft. 
  • Traditional drafting: Danielle was nearly done reviewing the first template brief. She was able to copy over some good articulations from that brief’s failure to warn section into her draft, but had also noted a lot of cases she would need to go back and cite-check as well as research paths to go down. 
See how far Jake and Danielle were able to get in 10 minutes through A.I. vs. a traditional drafting process.


After just ten minutes of research and drafting time, both Jake and Danielle had a document started with an outline of at least one argument. However, thanks to the power of A.I. and NLP in legal tech, Jake was able to build out much more of his draft in less time. Whereas Danielle was left with a number of research projects — including cases to cite-check and citations to update with more recent cases — Jake was able to quickly and confidently add up-to-date law to his draft. 

This head-to-head challenge was part of our recent webinar, “A.I. for Products Liability Motions.” You can watch the recording of the full webinar here. For more on how to write Products Liability briefs efficiently and effectively, tune in to this week’s webinar: https://compose.law/pl-draft-high-quality-briefs/