What AI Can — and Can’t — Do in Immigration

An O-1A or EB-1A petition can run to several hundred pages: a legal brief, recommendation letters, press, publications, contracts, financials, translations and an exhibit index that ties it all together. Much of the time spent on a petition goes not to legal argument, but to organizing material.

That is exactly where AI is useful. It is also where the temptation to let it do too much begins.

What AI does well

• Organizing evidence. Sorting hundreds of documents, extracting dates, names and figures, and grouping them by the legal criterion they support.
• Mapping evidence to criteria. Highlighting which criteria look strong, which are thin, and what is missing — so the strategy conversation starts from a complete picture.
• First drafts. Producing structured drafts of exhibit lists, cover letters and sections of the brief, built from the client’s own documents.
• Consistency checks. Catching mismatched dates, names spelled two ways, exhibits referenced but not attached — the small errors that undermine credibility.

What AI should not do

• Choose the strategy. Which category, which criteria to lead with, what to leave out, whether to file now or wait — these are judgments with consequences for a person’s life.
• Give legal advice. Advising someone on their immigration options is the practice of law. It belongs to a licensed attorney who owes duties to the client.
• Assert facts it can’t verify. Language models can produce confident, fluent text that is wrong. In court filings, that has already led to sanctions for lawyers who submitted AI-invented case citations.
• Sign or file. A petition carries legal attestations. The person who signs it must have read it, understood it and be accountable for it.

The professional rules already apply

In July 2024, the American Bar Association issued Formal Opinion 512 on lawyers’ use of generative AI. Its message is that the existing rules of professional conduct apply in full: lawyers must understand the tools they use well enough to use them competently, protect client confidentiality when information is entered into them, communicate with clients about their use where appropriate, supervise the output, and charge reasonable fees.

We think that framing is right. AI does not create a new kind of legal service. It changes how the work behind a legal service gets done.

How this works in a Milestone petition

• Intake. You upload your CV and documents. AI extracts and organizes the evidence.
• Mapping. AI maps the evidence to each legal criterion and flags gaps. Your attorney reviews the map and sets the strategy.
• Drafting. AI produces a first draft from your documents. Your attorney rewrites, strengthens and decides what the argument is.
• Review and filing. You review and approve. Your attorney signs and files.

Why it matters for clients

The benefit is not that a machine writes your petition. It’s that the attorney’s time goes where it matters most — on strategy and argument, not on assembling exhibits. That is what makes a fixed fee and a clear timeline possible without cutting corners.

Used this way, AI makes careful legal work faster. Used as a substitute for that work, it makes mistakes faster. The difference is the attorney.