Related Posts
Additional Posts in Law
New to Fishbowl?
Download the Fishbowl app to
unlock all discussions on Fishbowl.
unlock all discussions on Fishbowl.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Download the Fishbowl app to unlock all discussions on Fishbowl.
Copy and paste embed code on your site

Scan your QR code to download
Fishbowl app on your mobile

The AI assisted Westlaw works product quite well for for giving you citations to precedent when you already know the law but need the citations, provided you frame the questions right. I'm referring to the new one they're rolling out for big firms, not the one designed for small firms.
Two reason come to mind. First, in-house is not expected to be perfect, while law firms need to deliver consistent high caliber work product. Most legal GenAI tools allow someone to be decent at an area of the law they are not familiar with, but can be frustrating for lawyers in firms where it is what they do every day and need much more than just decent results. The second reason is incentives. In-house is constantly under pressure to be efficient and do more with less, so they may be more likely to dig in and figure out how to get the most out of the tools. Law firm lawyers lose billable time when they experiment with tools, so they are more likely to just develop basic skills and may not be using the tools correctly/in the best way.
Have you ever worked in-house? The standard is, good enough and risk aware. Our most used tool is Google. We are not worrying about perfection and typos to the absurd extent firm lawyers do.
Westlaw/lexis AI have a place. I'd ballpark them saving me 1-3 hours per week. It's another research tool basically.
The legal specific tools... Hit or miss. I build my own datasets and my sense is that they're basically a combination of some very light technical engineering mixed with some tailored datasets.
I mainly use perplexity pro and Claude, coupled with relevant materials and/or datasets, and use python to build datasets and scrapers
these are equivalent of a calculator or 3d modeling software for an engineer - useful tools at the hands of someone competent who can input the correct data and has enough insight to decipher the signal from the noise
Harvey keeps raising more money, so it must be working for somebody.
But really, in-house, we work as much or as little as we need to to get the job done. Efficiency gains go to our free time.
Firm lawyers bill by the hour. Efficiency just increases the amount of work.
I have found the Westlaw tool to be decent, but not all the way there yet. As someone else mentioned, if I know the point of law I am looking for (or if I know it exists), then it will usually find it and give me the cite. If I ask it something more open ended, it either won’t give me an on point answer or give me a dead wrong answer.
Examples:
1. Asked Westlaw AI whether a certain type of claim is tried to a jury or the bench. It found the exact case with the answer. Saved me 30 minutes of research.
2. Asked Westlaw if a certain type of motion required a particular type of notice. This was a yes/no question. It gave me the wrong answer. Fortunately, Westlaw always provides citations to the authority that it is relying on, so I was able to catch the error. Cost me 30 minutes of research confirming that the AI was wrong.
I am starting to think that the real strength of AI tools is that they are both fast and iterative. If the first draft isn’t right, you can keep sending it back to revise. After 45 minutes you get something a junior associate or law clerk might need 3 days to do. Maybe still not ready for prime time, but way further along than you might otherwise be.