The read is the work
AI almost put a mistake in front of an immigration judge in one of my client's filings. I caught it on a manual read.
The detail was small. My client spoke a particular version of English. My mentor had passed along a different version. The AI tool I was using to help assemble the filing kept reproducing the wrong version in the brief — even after I corrected it. The case had roughly 1,000 pages of material and a tight timeline. The tool would not self-correct. The only way to catch it was to read every page.
So that is what I did.
The AI's mistake becomes my mistake the moment it reaches the judge. The client does not get to explain that the tool got something wrong. The record is the record.
The rule
I read every page through.
I verify every reference, even the ones that look right.
I catch the hallucinations the tool cannot catch itself.
In immigration work — and especially in asylum work — the AI's mistake becomes my mistake the moment it reaches the judge. The client does not get to explain that the tool got something wrong. The record is the record.
I use AI in my practice. It saves real time. But the time it saves is not the time I would have spent reading the file. That time is the work.
Attention to detail in this practice is not a professional virtue. It is what the client is paying for, and it is what the system requires. There is no version of this job where the read gets skipped.
If you practice in this space — or work alongside lawyers who do — I would be curious how you are handling AI in your own work. Comments are open below.
Aakash Sharma is a Connecticut-licensed attorney and the founder of the Law Office of Aakash Sharma, LLC. His practice focuses on family-based immigration, asylum, naturalization, and K-1 fiancé visas.
Comments
There are no comments for this post. Be the first and Add your Comment below.
Leave a Comment