How to Use AI to Draft an Employment-Based Petition

If your firm handles employment-based immigration, you already know the math: a single EB-1A petition can take 25–40 attorney hours to draft to first-pass quality, an EB-2 NIW runs 15–30 hours, and even a "routine" H-1B specialty occupation petition burns 4–8 hours by the time the support letter, LCA reasoning, and evidence brief are tight. Learning how to use AI to draft an employment-based petition is the highest-leverage workflow change a business immigration practice can make this year.

This guide walks through the same workflow we teach at Drafty AI across every employment-based visa category USCIS recognizes — immigrant petitions (EB-1, EB-2 NIW, EB-3, EB-4, EB-5) and the popular non-immigrant categories (O-1, L-1, TN, E-2, E-3). It is written for practicing immigration attorneys, not generalists. By the end, you will have a repeatable, visa-specific drafting process you can apply on Monday morning.

What Counts as an "Employment-Based Petition"

The phrase covers two distinct universes of casework that share a common architecture: an employer (or self-petitioning beneficiary) is asking the U.S. government to authorize work, residence, or both, on the basis of the beneficiary's qualifications, the employer's needs, or both.

Immigrant employment-based categories (EB-1 through EB-5)

These lead to a green card. The petition is typically filed on Form I-140 (EB-1, EB-2, EB-3), Form I-360 (EB-4), or Form I-526 / I-526E (EB-5). The categories are:

• EB-1A — Extraordinary Ability. Self-petitionable. The beneficiary must establish sustained national or international acclaim under 8 C.F.R. § 204.5(h), either via a one-time major achievement (Nobel-level) or by satisfying at least three of ten regulatory criteria.
• EB-1B — Outstanding Professor or Researcher. Employer-petitioned. Requires three years of qualifying experience and satisfaction of at least two of six regulatory criteria under 8 C.F.R. § 204.5(i).
• EB-1C — Multinational Executive or Manager. Employer-petitioned. Requires one year of qualifying employment abroad in an executive or managerial capacity with a related entity within the three years preceding the petition.
• EB-2 NIW — National Interest Waiver. Self-petitionable. Governed by Matter of Dhanasar, 26 I&N Dec. 884 (AAO 2016), which sets a three-prong test: substantial merit and national importance of the proposed endeavor; the beneficiary is well positioned to advance it; and, on balance, it is beneficial to the United States to waive the job-offer and labor-certification requirements.
• EB-3 — Skilled Workers, Professionals, and Other Workers. Employer-petitioned and anchored on a certified PERM labor certification. Distinguished from EB-2 primarily by the educational and experience floor of the position.
• EB-4 — Special Immigrants. Filed on Form I-360. Covers religious workers, certain broadcasters, Afghan and Iraqi translators, special immigrant juveniles, and a long list of statutory categories under INA § 101(a)(27).
• EB-5 — Immigrant Investor. Requires a qualifying investment ($800,000 in a targeted employment area or $1,050,000 otherwise, post–EB-5 Reform and Integrity Act of 2022) and creation of at least ten full-time U.S. jobs.

Non-immigrant employment categories (work visas)

These authorize work for a limited period and a specific employer. Most are filed on Form I-129; TN and E categories can also be sought directly at a U.S. consulate or, for Canadian TNs, at a port of entry. Some of the most popular categories include:

• H-1B — Specialty Occupation. Requires a certified Labor Condition Application (LCA) and a position that normally requires at least a bachelor's degree in a specific specialty, under 8 C.F.R. § 214.2(h)(4)(iii)(A).
• L-1A / L-1B — Intracompany Transferee. L-1A for executives and managers, L-1B for specialized knowledge. Requires one year of qualifying employment abroad with a related entity within the three years preceding the petition.
• O-1A / O-1B — Extraordinary Ability or Achievement. O-1A for sciences, education, business, or athletics; O-1B for arts and motion picture or television production. Each has its own regulatory criteria under 8 C.F.R. § 214.2(o).
• TN — USMCA Professional. Available to Canadian and Mexican citizens in the professions listed in USMCA Appendix 2 to Annex 16-A (formerly NAFTA Appendix 1603.D.1). Education and credential requirements vary by profession.
• E-2 — Treaty Investor. Requires substantial investment in a real, operating U.S. enterprise by a national of a treaty country, with the enterprise more than marginal and the investor in a position to develop and direct it.
• E-3 — Australian Specialty Occupation. Functionally similar to H-1B for Australian nationals, with its own LCA requirement and a separate annual cap.

For most attorneys, "drafting the petition" means producing four things in parallel: the form itself, an employer or self-petitioner support letter, a legal argument brief (sometimes called the cover letter or memorandum of law), and an exhibit list. AI dramatically accelerates the last three.

Anatomy of an Employment-Based Petition — What AI Is Actually Drafting

Before walking through the workflow, it helps to be specific about what an AI tool is producing on an employment-based petition. AI is not filling out Form I-140 or I-129 for you — that lives in your case management system and depends on data you already have. AI is doing the heavy paragraph-by-paragraph work in three documents:

1. The support letter (employer letter or, for self-petitions, the petitioner statement). Tailored to the beneficiary's role, qualifications, and the visa category's regulatory criteria.
2. The legal brief that explains how the record satisfies each statutory and regulatory element, citing exhibits as it goes.
3. The exhibit list with descriptive captions that map to the brief.

How AI Changes the Drafting Workflow

What changes when AI drafts the petition is not the legal work — it is when the legal work happens. The case theory, criteria analysis, and evidence map still have to exist. They just have to exist in a form the model can read before it drafts, rather than living in your head as you write.

That shifts the workflow from "think while you draft" to three distinct phases:

Phase 1 — Structure the input

The judgment that used to happen during drafting now happens before it. The case theory, the regulatory criteria you intend to argue, the exhibits you have, and the comparable-evidence framing for any criterion that does not map cleanly — all of it goes into the prompt as structured input. A thin prompt produces a thin draft. A complete prompt produces a draft you can edit in one pass. For prompt-engineering principles applied to legal work, see our free AI prompting guide for immigration lawyers.

Phase 2 — Verify the output

Reviewing an AI draft is not editing. Before you touch prose, you are checking that the model addressed every criterion the category requires, cited the controlling regulation correctly, and referenced exhibits accurately. Hallucinated citations and invented facts are the failure modes that matter — typos and tone come after. Structural correctness first; prose second.

Phase 3 — Layer in strategy

The case theory, the rhetorical choices, the why-this-beneficiary-now narrative — that is where attorney value lives, and it is what AI cannot supply. The draft you get back is a competent skeleton. The strategic layer is yours alone, and it is the step that turns an AI draft into an attorney work product.

Visa-by-Visa Playbook: How to Prompt AI for Each Employment-Based Category

Each category has its own regulatory architecture, and the right AI prompt mirrors it. The notes below are not magic strings — they are the structural shape of a working prompt. Adapt them for each case.

EB-1A — Extraordinary Ability

EB-1A is the highest-stakes self-petition in employment-based immigration. The regulatory test under 8 C.F.R. § 204.5(h)(3) asks for satisfaction of at least three of ten criteria (awards, memberships, published material about the beneficiary, judging, original contributions, scholarly articles, exhibitions, leading or critical role, high salary, commercial success) plus, under the Kazarian two-step, a final-merits determination of sustained acclaim.

The two AI tasks that dominate an EB-1A draft are criterion-by-criterion argumentation and comparable evidence framing for criteria that do not map cleanly. A well-structured prompt names the three or more criteria you intend to argue, lists the exhibits for each, and specifies the comparable-evidence framing for any criterion you cannot meet on its face.

Common AI failure mode: defaulting to a generic "the beneficiary has received awards" paragraph that could apply to anyone. Force specificity by listing the awards by name, the granting body, the field of competition, and the selection process in the prompt itself.

EB-1B — Outstanding Professor or Researcher

EB-1B requires three years of qualifying experience plus two of six regulatory criteria under 8 C.F.R. § 204.5(i)(3): major prizes, memberships requiring outstanding achievement, published material about the beneficiary's work, judging, original contributions, or scholarly articles. The employer must also offer a permanent research position or tenure-track teaching role.

The AI workflow mirrors EB-1A but is anchored on the employer side. The strongest prompts include the employer's research mission statement, the position description, and the beneficiary's role within ongoing funded research. Citation counts, journal impact factors, and grant award histories belong in the prompt input as structured facts, not adjectives.

EB-1C — Multinational Executive or Manager

EB-1C is the green-card analog of L-1A. The two regulatory hinges are qualifying employment abroad (in one of the three years preceding the petition in a managerial or executive capacity with a related entity) and continuing executive or managerial capacity in the U.S. role.

AI is strong at writing the organizational-structure narrative: who reports to whom, what decisions the beneficiary controls, what budget authority they hold, what strategic direction they set. Feed the org charts and decision matrices into the prompt as structured data. AI is weak at the qualitative judgment of function vs. first-line supervision, which is the most common ground for an EB-1C denial. That distinction is an attorney call.

EB-2 NIW — National Interest Waiver

Matter of Dhanasar, 26 I&N Dec. 884 (AAO 2016), governs every petition filed after January 2017 and sets a three-prong test:

1. The proposed endeavor has both substantial merit and national importance. The endeavor — not the beneficiary — is the subject of the first prong.
2. The beneficiary is well positioned to advance the endeavor. Evidenced by education, skill, record of success, and a credible plan forward.
3. On balance, it would be beneficial to the United States to waive the job offer and labor certification requirements.

An NIW prompt should be organized around these three prongs in order. The most common AI failure mode is conflating the beneficiary with the endeavor — every NIW prompt should explicitly define the endeavor in the first sentence and then describe how the beneficiary advances it.

EB-3 — Skilled Workers, Professionals, and Other Workers

EB-3 petitions are anchored to a certified PERM labor certification. The drafting work is less about novel legal argumentation and more about showing that the beneficiary meets the minimum requirements of the certified PERM and that the employer has the ability to pay the offered wage. AI is genuinely useful for the support letter and the ability-to-pay narrative; it is a poor fit for the PERM itself, which is a DOL process governed by ETA Form 9089 and not a drafting exercise.

EB-4 — Special Immigrants

EB-4 is a heterogeneous category. The most common subcategories are religious workers (often R-1-to-EB-4 conversions), special immigrant juveniles, and Afghan or Iraqi translators. Each has its own evidentiary framework. AI is most useful inside subcategories that have a stable evidentiary pattern — religious worker petitions in particular, where denomination membership, compensation, and bona fide employer questions are pattern-rich. Use category-specific prompts; do not try to draft "an EB-4 petition" generically.

EB-5 — Immigrant Investor

EB-5 is the most document-intensive employment-based category. The two legal hinges are lawful source and path of funds and job creation. AI dramatically accelerates the source-of-funds narrative — tracing capital from earnings, gifts, loans, or asset sales through bank statements, tax returns, and asset documentation. AI is also strong on the business plan summary that explains how the investment will produce ten qualifying U.S. jobs.

The risk in EB-5 drafting is fabricated facts: general-purpose AI tools occasionally fill gaps with plausible-sounding but unsupported financial claims. Every dollar figure in an EB-5 brief must trace to an exhibit. Use a specialized tool, and verify aggressively.

O-1A and O-1B — Extraordinary Ability or Achievement

O-1A (sciences, education, business, athletics) and O-1B (arts, motion picture and television) are the most criteria-driven non-immigrant petitions in U.S. immigration law. The regulatory framework at 8 C.F.R. § 214.2(o)(3)(iii) for O-1A and (iv) for O-1B is closely parallel to EB-1A in structure — multiple criteria, comparable evidence for criteria that do not map, a final merits determination.

Two practical notes. First, USCIS issued an October 2021 policy memo, with successor updates, clarifying how STEM-field evidence maps to the O-1A criteria; current prompts should reference that policy guidance explicitly. Second, an O-1 almost always requires a written consultation from an appropriate peer group or labor organization. AI can draft the consultation request; only the issuing organization can produce the consultation itself.

L-1A and L-1B — Intracompany Transferee

L-1 petitions hinge on three documentary questions: qualifying relationship between the foreign and U.S. entities (parent, subsidiary, branch, or affiliate); qualifying employment abroad for one continuous year in the three years preceding the petition; and qualifying role in the United States (executive or managerial for L-1A; specialized knowledge for L-1B).

AI is strong on the corporate-structure narrative and the duties description. L-1B "specialized knowledge" is the harder draft — it requires demonstrating that the beneficiary's knowledge is distinct from what is generally found in the industry. Specialized-knowledge prompts should include concrete examples of the proprietary processes, products, or methodologies the beneficiary has mastered, not adjectives.

TN — USMCA Professional

TN status is available to Canadian and Mexican citizens working in one of the professions listed in the USMCA Appendix 2 to Annex 16-A. The drafting deliverable is short — typically a one-to-three-page employer support letter plus credential evidence — but the legal traps are well known. The position must fall within a listed profession, the duties must be professional in nature, and the beneficiary's credentials must match the profession's specific requirements (a Canadian engineer must hold a baccalaureate or provincial licensure; a management consultant must show the specific experience and credentials required for that role).

AI is useful for the support letter and the matching-of-duties-to-profession narrative. It is not useful for the credential evaluation, which often requires a foreign-degree evaluator.

E-2 — Treaty Investor

E-2 cases are evidentiary marathons. The five recurring questions are treaty country nationality, substantial investment, real and operating enterprise, more than marginal, and develop and direct. AI accelerates the business plan summary, the marginality analysis, and the source-of-funds narrative. For E-2 prompts, include the specific industry, the investment amount, the proportionality argument (investment relative to the cost of establishing a similar business), and the five-year job creation projection.

E-3 — Australian Specialty Occupation

E-3 looks like H-1B and drafts like H-1B, with two important differences: the LCA is filed for an E-3 position rather than H-1B, and the position itself must require theoretical and practical application of a body of specialized knowledge requiring a bachelor's or higher in the specific specialty. Re-use H-1B specialty occupation prompts; swap the LCA references and add the Australian-citizenship documentation.

What AI Still Cannot Do on an Employment-Based Petition

The list is short but non-negotiable.

• Choose the category. EB-1A vs. EB-2 NIW vs. EB-1B is a strategic decision based on the beneficiary's record, priority-date pressures, and risk appetite. That is attorney work.
• Read the client. Credibility, candor, and the small inconsistencies in a beneficiary's record that telegraph trouble are picked up in client interviews, not in document review.
• Decide what evidence to gather. AI drafts around the evidence that exists. The strategic call to commission a new expert letter, to seek a recommendation from a particular peer, or to skip a weak exhibit altogether is the attorney's.
• Sign the brief. Your bar number, your professional responsibility, your reputation.

For the ethical framing of AI use in legal practice, the ABA's guidance on AI and the Florida Bar's AI ethical guidelines are the right starting points.

Common Mistakes When Drafting Employment-Based Petitions With AI

Citing the wrong regulation or generation of policy guidance

General-purpose AI can pull older or outdated citations. The Matter of Dhanasar three-prong test replaced Matter of New York State Department of Transportation for NIW petitions in 2016 — but generic AI tools sometimes still produce briefs framed under the old test. Specialized immigration AI tools that are trained or grounded on current adjudication patterns substantially reduce this risk. Verify every citation against the source before filing.

Skipping comparable evidence framing

For EB-1A, EB-1B, and O-1 petitions, "comparable evidence" is the regulatory pressure valve that lets you argue alternative proof for criteria the beneficiary cannot meet on their face. AI tools that do not understand comparable evidence will mark a criterion as unmet when it could have been argued. Prompt for comparable-evidence framing explicitly.

How Drafty AI Makes Employment-Based Drafting Easier

Everything in this guide — the category-specific criteria, the evidence-to-criterion mapping, the regulatory framing — is the work we built Drafty AI to handle for you. The reason most attorneys eventually move off general-purpose tools is that the prompt itself becomes the job: writing it, refining it for each case, keeping it current with regulatory and policy changes. Drafty AI removes that step entirely.

Pick the case type. Upload the evidence. That's it.

You tell Drafty AI what you are drafting — EB-1A, EB-2 NIW, EB-1C, O-1A, L-1B, TN, E-2, whatever the category — and upload the evidence pile. There is no prompt to write. The platform already knows the regulatory criteria for the category, the comparable-evidence framing where it applies, the controlling case law (Matter of Dhanasar for NIW, Kazarian for EB-1A and O-1), and the USCIS-conforming output structure adjudicators expect. The category is the prompt.

Trained on immigration practice, not the open internet

Drafty AI was built specifically for immigration attorneys. It has been trained on the regulatory frameworks, USCIS adjudication patterns, and petition structures that govern employment-based work — the things general-purpose AI tools have to be told case by case. Updates to policy guidance, new precedent decisions, and regulatory shifts are reflected in the platform, so you do not have to maintain your own prompt library or worry about an outdated test sneaking into a draft.

What you get back

A draft brief, support letter, and exhibit list in USCIS-conforming format, with citations in the style adjudicators expect, evidence mapped to criteria, and gaps flagged where evidence is missing. The attorney work is what it should be — strategy, judgment, sign-off — instead of prompt engineering.

Frequently Asked Questions About AI for Employment-Based Petition Drafting

How much time does AI save on an EB-1A or EB-2 NIW petition?

The consistent pattern across attorneys using AI for employment-based drafting is a 60–75% reduction in first-draft time. An EB-1A petition that historically took 25–40 attorney hours typically moves to 7–12 hours, concentrated in evidence gathering, strategy, and final attorney review. An EB-2 NIW often drops from 15–30 hours to 5–10. Across a business immigration practice, the savings compound quickly. For practice-wide time savings, see our guide to 113 time-saving tips for immigration attorneys.

Is it ethical to use AI for employment-based petitions?

Yes, with the usual duty-of-competence framing: the attorney must understand what the tool does, verify its output, protect client confidentiality, and retain decision authority. The ABA and the Florida Bar have both published AI guidance on these principles, and they are aligned. Using AI to draft a petition is professional practice; delegating attorney judgment to AI is not.

Which AI tool is best for employment-based petition drafting?

For the structural answer, see our comparison of ChatGPT vs Claude for immigration lawyers, which also covers where specialized immigration AI tools fit. The short version: general-purpose AI works for brainstorming and outlining; specialized immigration AI works for production drafting where USCIS-conforming output, regulatory accuracy, and confidentiality matter.

Can I use AI for the response if an employment-based petition draws an RFE?

Yes — that is the workflow covered in our companion guide, How to Use AI to Draft an RFE Response. The petition and RFE workflows are deliberately structured to share inputs: the evidence inventory, the criterion-by-criterion mapping, and the regulatory framework carry over directly.

How do I try Drafty AI for free?

Start a free trial at app.draftyai.com/signup — no credit card required. You get full access to draft employment-based petitions across every category covered in this guide.