Where legal AI breaksFour problems, and what to do about each.
Written from what lawyers say goes wrong in public, not from a feature list. Start with the one that sounds like your week.
Start here
Pick the problem you have.
You are doing the same work twice
One matter workspace keeps the case files, the research and the drafts in place, so getting the case in and the draft out does not start over every session.
You still have to check what the AI wrote
If you still have to check everything an AI drafted before you can trust it, the tool did not save you the time it promised. Here is the workflow that gets the check itself down to minutes, not a redo.
Your firm will not approve an AI tool
You are pushing AI adoption at a 10 to 50 lawyer firm. This free packet answers what your managing partner, the person who runs the firm and signs off, will ask about client data, so you forward it once instead of defending it call by call.
Someone else's AI draft landed on your desk
Work arrives already drafted by AI, from clients, pro se opponents, opposing counsel. Here is how to triage it and how to bill for it.
You are doing the same work twice
Never Start From Scratch
One matter workspace keeps the case files, the research and the drafts in place, so getting the case in and the draft out does not start over every session.
Document too long to process
The context window caps how much it holds, so the agreement goes in piece by piece. Work it whole instead.
Copy-pasting into Word
Drafting in a chat means moving it to Word and fixing what broke. Draft and revise inside Word itself.
Recreating the same template
Rebuilding the clauses and the formatting from the last matter every time. Reuse it with the formatting intact.
Re-explaining the case
A new session does not know the facts of the case. The matter keeps them, so you do not repeat them.
Uploading the same files
The same case files uploaded at the start of every session. They stay in the matter for next time.
Work fragmented across tools
Research in one tool, drafting in another, checking in a third, and you carry the case between them. All three in one place.
You still have to check what the AI wrote
The Verification Tax
If you still have to check everything an AI drafted before you can trust it, the tool did not save you the time it promised. Here is the workflow that gets the check itself down to minutes, not a redo.
Checking citations yourself
A free, step-by-step way to check an AI-drafted citation yourself.
Sanctioned for fake citations
Nine documented court sanctions for fabricated AI citations.
Sourced or just confident
Why some AI answers are sourced and others are confident guesses.
The babysitting problem
What it costs you when a draft is not built to be checked.
Your firm will not approve an AI tool
The Managing Partner Packet
You are pushing AI adoption at a 10 to 50 lawyer firm. This free packet answers what your managing partner, the person who runs the firm and signs off, will ask about client data, so you forward it once instead of defending it call by call.
The training question
The one question that stalls AI adoption, answered plainly.
Security back and forth
The standard vendor questionnaire, pre-answered for Irys.
The policy nobody follows
Why written policy alone does not stop the workaround.
When the partner says no
Five objections, honest answers, and a low-stakes pilot.
Someone else's AI draft landed on your desk
The AI Mess
Work arrives already drafted by AI, from clients, pro se opponents, opposing counsel. Here is how to triage it and how to bill for it.
In their own words
What lawyers say about all four.
29 verbatim posts from public forums where lawyers talk about their work, sorted into the same four problems.
See Irys on your own matter.
Tell us about your practice and we will set up a walkthrough on the work you actually do.
No credit card. Transparent pricing. 7-day free trial.