AI for O-1, EB-1, and NIW Applicants (Part 2)

Part 2: Working With an AI-Savvy Attorney

Why “AI-savvy” matters right now

An AI-fluent attorney produces stronger first drafts, runs more rounds of revision in the same calendar time, and has more bandwidth left for strategy. Note the number of human hours on the parts that require client communication, analysis and judgment should remain the same, or even increase. The use of AI should improve the quality of the drafts, the accuracy of the final petition, and the efficiency of the client-attorney collaboration.

Ask your attorney how they use AI

Open the conversation directly, during intake. These are fair questions and any competent practitioner should have ready answers:

• Which AI tools do you use, and for which tasks?
• How is my data handled? Paid or enterprise tiers, no-training settings, retention policies.
• How do you review and verify AI-assisted drafts?
• What does AI do and not do on a case?

On the last question, here is how the division of labor should look:

Attorney work, not AI work: strategy, criteria analysis, evidence sufficiency, final judgment, RFE risk assessment.

Appropriate AI tasks: first rough drafts, summarizing, proofreading, consistency checks.

Mixed: research, where AI surfaces sources and the attorney verifies and synthesizes them.

If an attorney tells you they use no AI at all, that is a legitimate practice choice. If they do use AI, you are entitled to know how.

Do not send your attorney AI slop

This is the single most important rule of client-attorney collaboration in my firm.

By AI slop I mean machine-generated text you forward to your lawyer without working on it: bloated, generic, repetitive, and full of plausible-sounding claims that nothing in your record supports.

Why not? Because you will either pay a lot more money (if you pay hourly), or you will just make your attorney’s work a lot harder for no reason. It takes your attorney much longer to read, fact-check, and untangle AI slop than to just write the section or the letter from scratch. You are wasting everyone’s time and consuming attorney attention that should have gone to strategy.

Four practical rules keep this from happening.

Ask what your attorney actually needs from you. Every firm's workflow is different. Some want a short factual intake with no AI involvement at all (that’s our firm’s general preference). Some want a rough draft in your own voice that they will develop further with AI. Some are happy to receive a full AI-assisted draft. Work inside your lawyer’s workflow.

Never forward raw output. Read every line. Ensure full accuracy. Cut anything you cannot personally confirm. Cut repetitions. Cut padding and verbiage that don’t add anything to the substance.

Catch AI hallucinations. If the AI invents a fact or makes an unfounded assumption, you must catch it. It is your case, you know it better than anyone. Your attorney should be able to rely on facts you provide. A missed error may lead to last minute corrections and delays or, worst of all, end up in the filed case and put the entire petition at risk.

Explain what you send. Something as simple as highlighting a part you’re not sure about or adding a comment tells your attorney where to concentrate.

Build a shared evidence repository early

An organized folder, kept current, dramatically accelerates AI-assisted drafting on both sides. It should hold:

• Your CV, updated.
• Criteria-based evidence, with background information about awards, organizations, media, competitions, etc.
• Recommender names, bios, and a short summary of your relationship with each.
• Immigration history documents, if applicable.

It is important because an AI model that can see your entire record produces more complete and accurate drafts.

Ask for adversarial stress-testing

Before filing, ask your attorney to run an AI-driven inconsistency test across the draft petition and the exhibits, or run one yourself.

This is a high-value use of AI that most clients do not know to request. Dates that drift, job titles that shift between documents, exhibit numbers that do not match the table, factual claims in a recommender letter that the exhibits do not support: all of these can be identified and addressed before the filing, instead of being discovered by USCIS and drawing an RFE.

Watch for red flags

There are warning signs that AI is being used by the firm without enough human oversight and revisions:

• Recommender letter drafts that sound interchangeable.
• A personal statement so generic it could belong to anyone in your field.
• Citation errors or fabricated authorities.

Any of the three is a reason to slow down and go through the file carefully before it gets submitted.

Reserve strategy and judgment for your attorney

Your attorney's experience should generally prevail over an AI recommendation, especially in strategy questions:

• Which criteria to claim, and which to drop.
• How to characterize your field and your contributions.
• How to handle weak spots in the record.

These are judgment calls informed by experience and knowledge built from filed cases and adjudicator behavior rather than from text on the internet.

What we built for this

Because so much of the above comes down to structured collaboration, we built a tool for it: an AI-powered client collaboration application for O-1, EB-1, and NIW cases, including an intake questionnaire, a recommendation letter workflow and the final support letter review. It is designed to make our collaboration the most efficient, to keep your evidence organized from the start, and to produce letter drafts that cover your complete record. You can find it at our appointment page ____ or at app.lisaesq.com.