The American Immigration Lawyers Association and a company called Visalaw.ai recently announced the second generation of an AI tool for immigration lawyers called Gen.

According to Visalaw.ai’s website, the platform has three “dimensions”:

“Gen,” a research and summarization tool

“Draft,” called the “cornerstone project” of Visalaw.ai, a tool for creating complex immigration law documents

“Engage,” a chat feature to screen clients and suggest pathways of relief to the lawyer

Humanitarian Limits

Right now, Gen seems poised to serve mostly the market for employment-based immigration, a service usually provided by immigration departments of large law firms for their regular corporate clients.

The first version of Gen contained mostly employment-based material; beta testers found its answers to asylum and other humanitarian questions to be less useful because of a limited data set (mostly online government sources). Gen version 1.0 (really the second version) released last week happily draws from a more expanded library of sources on asylum and removal, according to Visalaw.ai founder and immigration lawyer Greg Siskind.

That’s great, but a subscription to Visalaw.ai currently costs $220 per user, per month, according to the company’s website. At that rate, few legal services organizations or law school clinics will be able to afford access. The website anticipates costs coming down with scale, so maybe someday Visalaw.ai will be accessible to the lawyers who help the clients with the greatest risk and the least support.

Right now, it seems like Gen can mostly help business immigration lawyers compete with other business immigration lawyers.

Does Generative AI Work for Immigration?

That’s assuming, of course, that Gen works. So far, initial experiments by me, my colleagues, and my students suggest that generative AI is, at best, about as good as a very fast but wildly unreliable research assistant. It can bring you sources and analysis, but since a huge percentage of it is wrong or even completely fictional, you’d better not rely on any of it. And you lose the learning that comes from doing the research yourself.

Perhaps generative AI will soon succeed at taking client documents as inputs and producing first drafts of technical but standard immigration documents (say, an employer’s letter petitioning for a green card for its employee, or an brief of asylum and withholding of removal in a particular circuit). Much of the law is boilerplate anyway; after verifying the factual information, the attorney could quickly determine whether the rule blocks and analysis make sense.

As a professor, however, I sense a danger that many lawyers may confused efficiency with advocacy. Particularly in the removal context, where the attorney makes arguments to move the judge, the human element of communication goes far beyond crafting accurate rule blocks and plugging in standard analyses. As we try to impress upon students in the Immigration Clinic, advocacy is not just telling a story, it’s crafting a narrative. The difference between the two is substantial, and essentially human.

A good advocate could still use a sophisticated draft from a generative AI tool as a starting point, saving time to focus on the human elements: Client communication, empathy, reading the court’s temperament, crafting moving and powerful narrative. If generative AI nudges attorneys to focus more on these human skills, it will be a win for immigration lawyering. If it instead lulls attorneys into treating cases like widgets to run through a generative assembly line, it will be a big loss.