2 Comments
User's avatar
Mark Stansbury's avatar

Automated tagging has been painfully inaccurate in the past—good enough to be promising, but bad enough to be unreliable. I hope they’ve got it mostly figured out.

I’ve also found AI useful for mundane “core” tasks like writing the first draft of letters (cover letters, demands) and I’m testing it out with routine board consents and that sort of thing.

Pro tip: if you need to use names and confidential details, tell the AI to use CLIENT ONE or SETTLEMENT DOLLARS and replace those easy to spot placeholders in the final output.

Expand full comment
Zach Abramowitz's avatar

Interesting RE: past historical performance of automated tagging. LLMs for tagging though should be a game changer because they can look for meaning and get enough right to then make the traditional searches (which are still less expensive than searching with AI) much more impactful.

Expand full comment