Why immigration needs its own AI playbook
Immigration is the practice area where AI can pay back its cost in a single I-589. A solo attorney with thirty active cases moves thousands of pages a month between client, government, and file.
That document load is the reason AI matters here. Auto-populating a Form I-130 from a clean intake sheet saves an hour per filing. A summary of a 90-page Country Conditions package saves an evening. A first draft of an RFE response tied to the specific notice language saves half a day. Multiply that across a 50-case docket and the math gets serious.
The catch sits on the other side of the ledger. Immigration files contain some of the most exposed data in any law office: A-numbers, USCIS receipt numbers, country of origin, asylum claims, prior removal orders, sealed juvenile immigration records, and in some cases the client’s location. Paste the wrong PDF into a free chatbot and you have not only walked through Rule 1.6, you may have created a discoverable record on the wrong side of a DHS server.
This guide is for the solo or small-firm immigration attorney (1 to 10 lawyers, plus the paralegals and legal assistants who actually run the workflows). Family-based and employment-based practice is the default; removal defense and asylum get called out where the analysis changes. For the broader view across practice areas, see the pillar on AI for law firms. For the firm-wide rules of the road on AI use, see AI policy for law firms.
The five workflows worth setting up first
Build one workflow this quarter: form preparation. Time for two? Add USCIS policy tracking. The other three pay back fast once those two are running.
1. Form preparation from intake data
The use case: take a structured intake (passport bio page, marriage certificate, prior immigration history, employment record) and auto-populate the relevant USCIS forms. I-130 and I-485 for adjustment of status, I-129 and I-140 for employment-based petitions, I-765 for work authorization, N-400 for naturalization, I-589 for asylum. This is the most-cited use case in vendor marketing and also the one where a small firm sees the fastest return.
Immigration-specific platforms beat general-purpose chatbots for this work. Docketwise, Clio Duo, Visalaw AI, Imagility, and AILaw all maintain current versions of USCIS forms and map intake fields to form fields without a prompt. A generalist model can do the same work but you have to feed it the blank form each time and verify edition dates against the USCIS site.
Sample prompt for the generalist case, run inside an enterprise tier with a no-training agreement:
You are preparing a [FORM NUMBER, e.g. I-130] petition for [PETITIONER NAME] on behalf of [BENEFICIARY NAME]. I am attaching the intake worksheet, the petitioner's passport bio page, the marriage certificate (if applicable), and the prior immigration history summary. Produce: 1. A completed Form [FORM NUMBER] (latest edition per USCIS) with every field filled from the attached documents. Use [TBD] for any field that the attachments do not answer. Do not invent data. 2. A cover sheet listing every supporting document referenced in the petition, in the order it should appear in the filing. 3. A checklist of the fee, signature, and supporting-document requirements per the current USCIS instructions for this form edition. Verify the form edition date against the USCIS instructions page. If the attached blank form is not the current edition, stop and say so.
Verify every populated field against the source documents before filing. Models invent dates and reverse country fields more often than they admit.
2. USCIS policy and Federal Register monitoring
Policy churn is the second time sink in immigration practice. The USCIS Policy Manual updates several times a quarter. The Federal Register issues new rules and proposed rules. EOIR changes scheduling and filing practices. State immigration laws shift. A solo attorney cannot read every update and still bill.
Use a paid model with a long context window to summarize a week’s worth of updates into a one-page brief: what changed, which clients it affects, what action you have to take this week. Visas.ai and Imagility both market real-time policy alerts as a feature; if the budget is there, those tools save you the manual feed.
Sample prompt for the manual version, run on the week’s PDFs from USCIS and the Federal Register:
You are reviewing the attached USCIS Policy Manual updates and Federal Register notices published between [START DATE] and [END DATE]. Produce a one-page brief with these sections: 1. New rules and proposed rules: title, citation, effective date, one-sentence summary, the form or program affected. 2. Policy Manual updates: chapter and section, one-sentence summary of the change, the form or program affected. 3. Action items for our office this week: a numbered list. Tie each item to a specific client matter type (e.g. "pending I-485 cases need a refiled medical exam") rather than to client names. Cite the page or section number for every item. Do not paraphrase loosely; if the citation is to a specific paragraph, quote it.
Read the brief, then verify any item that would change a filing strategy by pulling the source. AI summaries of policy are starting points, not citations.
3. Document collection and client intake
Immigration intake is a paperwork relay. The client sends documents in three languages over six weeks. Half the documents are mislabeled. The other half are missing pages. A model that reviews each incoming PDF, identifies the document type, flags missing pages, and updates the case file checklist saves a paralegal a couple of hours a week per case.
Docketwise, Clio Duo, AILaw, and the eImmigration platform all have this built in for the most common documents (passport, marriage certificate, birth certificate, school records, employment letters, tax returns). For an unusual document or a non-Roman alphabet, a paid ChatGPT or Claude seat with vision can identify and translate without round-tripping to a human translator first.
Confidentiality rule: redact the A-number and the client name from the file name and metadata before any document goes through a generalist LLM. The redaction is fast and the audit trail it creates protects you under Rule 1.6.
4. RFE and NOID response drafting
A Request for Evidence is a structured prompt from USCIS. They tell you what is missing, what evidence they expect, and how much time you have. A language model is good at exactly this kind of structured response.
Feed the RFE, the underlying petition, and the supporting evidence into a long-context model and ask for a first draft of the response. The model will produce a cover letter, a section addressing each enumerated issue in the RFE, and an updated table of contents for the supplemental evidence package. Plan on 30 to 60 minutes of attorney revision on top of the first draft, which still beats writing the response from a blank page.
You are drafting a Request for Evidence response for [PETITIONER NAME] on Form [FORM NUMBER, e.g. I-140] for [BENEFICIARY NAME]. Inputs: 1. The RFE notice issued by USCIS on [DATE], with reference number [RFE REFERENCE]. 2. The original petition and supporting evidence (attached). 3. New supplemental evidence we are submitting in response (attached). Produce: 1. A cover letter addressed to the USCIS service center identified in the RFE. 2. A response section for each numbered or bulleted issue the RFE raises. For each issue: restate the request in one sentence, then explain how the original record plus the supplemental evidence satisfies the request. Cite the specific exhibit and page number for every factual claim. Use the regulatory citation USCIS used in the RFE. 3. An updated table of contents listing every exhibit in the response package in filing order. Do not invent exhibits. If the supplemental evidence does not address an issue, flag the issue at the top of the response so the attorney can decide whether to ask for additional evidence from the client or to file a partial response.
Same rule as before: enterprise tier for unredacted matters, paid Team tier for fully redacted ones, citation verification on every regulatory reference. The Notice of Intent to Deny workflow is the same shape, with a tighter deadline and a sharper tone in the cover letter.
5. Country conditions research for asylum and humanitarian relief
I-589 preparation is the most evidence-heavy filing on the family-and-humanitarian side. Country Conditions packages run 100 to 300 pages of human rights reports, news articles, and expert declarations. A model that summarizes the Department of State’s annual Country Reports on Human Rights Practices, the relevant UNHCR materials, and three to five recent news pieces into a digestible brief is the fastest way to scope the evidence needed for a particular claim.
Use a paid Claude or ChatGPT seat with a long context window. Country Reports are public documents, so confidentiality concerns are lower for the research stage; the client-specific declaration is where redaction matters again.
You are scoping a Country Conditions package for an I-589 claim filed by a [NATIONALITY] [GENDER] [AGE] who fears persecution on the basis of [PROTECTED GROUND]. The relevant time period is [START YEAR] to present. Inputs: the latest U.S. Department of State Country Report on Human Rights Practices for [COUNTRY], the most recent UNHCR eligibility guidelines for [COUNTRY], and the attached news articles. Produce: 1. A two-page summary brief organized by the elements of an asylum claim (well-founded fear, nexus to a protected ground, government involvement or government inability to protect, internal relocation). Cite the page or paragraph of each source for every factual claim. 2. A list of the strongest three to five pieces of evidence in the attached materials, with one-sentence relevance notes. 3. A list of gaps in the evidence: what is missing for this specific claim, and what document type would close the gap. Do not invent citations. If a fact does not appear in the attached materials, say so.
This workflow does not replace the asylum officer interview prep or the declaration drafting, but it saves the half day usually spent assembling the package outline.
Tools to know in 2026
Immigration-specific platforms
Docketwise remains the most-used case management system in family-based and employment-based immigration. Pricing is per-attorney per-month, mid-tier for small firms; the AI features (form auto-population, document review) are bundled in the higher plans.
Clio Duo is Clio’s AI layer on top of the broader Clio practice management suite. Good fit if you already run Clio Manage for billing and case management and want the AI features without switching platforms.
Visalaw AI focuses specifically on immigration document drafting and form preparation. Smaller player than Docketwise but priced for solos.
Imagility is strong for employment-based work (H-1B, L-1, PERM, I-140) and includes real-time policy alerts. Often used by firms with a heavy corporate immigration desk.
AILaw is a newer entrant. Worth a demo if you are starting from scratch and want a single platform for case management, form prep, and client portal.
Generalist tools that fit immigration work
For unredacted client material, use an enterprise tier with a written no-training agreement: Claude for Work, ChatGPT Enterprise, or Microsoft Copilot for Business.
For redacted or de-identified research, country conditions work, and drafting, a paid Claude or ChatGPT seat is the workhorse. CoCounsel is the bench option for litigation-heavy practices including BIA appeals and federal court immigration matters.
The caveats that matter most
Client confidentiality and the no-training agreement
The ABA’s Formal Opinion 512 applies Rule 1.6 to AI tools directly. Pasting an A-number and a country of origin into a free-tier LLM whose terms allow training is a disclosure. Use an enterprise tier with a written no-training agreement for unredacted matters. For everything else, redact the A-number, the client name, and the USCIS receipt number before the document leaves your office. The Catholic Legal Immigration Network (CLINIC) safety guide is worth reading regardless of firm type.
DHS exposure and the asylum file
Asylum claims are confidential under federal regulation, and disclosure to a third party can have consequences beyond a bar complaint. Treat I-589 declarations, country conditions tied to a specific client, and any communication referencing a pending asylum claim as the highest-sensitivity material in the office. Enterprise tier only. Redact before drafting. Keep an audit log of which document went through which tool.
Translation accuracy
Models translate well in common language pairs and miss in less-common ones. Every AI translation of evidence that will be filed goes through a human back-translation before it goes to USCIS. The certification of translation requirement under 8 C.F.R. 103.2(b)(3) still applies; an LLM is not a certified translator.
Citation verification
Read Mata v. Avianca if you have not. Two lawyers got sanctioned after filing a brief full of citations a chatbot invented. The same risk lives in BIA appeal briefs and federal court immigration filings. Every AI-drafted brief, motion, or RFE response goes through a citation check before it leaves your office. Pull the actual cases. If the model offers a quotation, find it in the opinion or strike it.
State bar and immigration court guidance
State bars keep issuing AI-use opinions. Florida, California, New York, and the District of Columbia bars have all published advisories. Some immigration courts have local standing orders on AI-drafted filings. Check both your bar and the specific court before a workflow turns into a habit. This article educates; it is not legal advice. Verify with your bar and your malpractice carrier.
Where to start this month
Week one, pick one open case and one paid LLM seat or one immigration-platform trial. Run the form-preparation workflow on a routine I-130 or I-765. Compare the output against the file you already know. Note what the model got right, what it got wrong, what it missed.
Week two, run the policy-monitoring prompt on the last 30 days of USCIS Policy Manual updates and Federal Register notices. Compare against the alerts you already receive (AILA, your CLE feed). If the brief catches something your existing feeds missed, you have your second workflow.
Weeks three and four, build a redaction protocol and a one-page firm policy. The protocol decides what you redact, who does it, and where the redacted file lives. The policy covers approved tools, redaction rules, citation verification, translation back-checks, and who reviews AI output before it leaves the office. The strategy article on AI policy for law firms has a template you can adapt.
Related on Business AI Workflows
If you want to keep going, these cluster articles cover the adjacent practice areas and the firm-wide rules that govern any AI rollout:
- AI for personal injury lawyers: the practice-area playbook for plaintiff PI firms, including medical chronology and demand letter workflows.
- AI for criminal defense lawyers: discovery review, suppression motions, and sentencing memos for state and federal criminal practice.
- AI policy for law firms: the policy template and rollout plan that supports every workflow in this article.
