Automation Case Study: Visa Document AI Automation Case Study in 3 Steps
If you're on a legal team that frequently handles overseas visa applications, you've probably faced this challenge at least once.
Beyond drafting applications, you need to verify various supporting documents, ensure compliance with the latest regulations, and complete submissions without issues.
Problems like review omissions, failure to reflect legal updates, and content inconsistencies in this process can lead to the critical consequence of visa rejection from just a single small mistake.
Moreover, when every review step is done manually, hours fly by just processing a single application.
As workload increases, errors become more frequent and time delays repeat.
The fatigue level of legal professionals in these situations is significant.
Repetitive Manual Work Can't Continue Like This
Recently, a company reached out to us.
"We need to manually check application drafts and dozens of types of supporting documents one by one, and each case takes at least 1-2 hours. As workload increases, mistakes multiply, and most importantly, reflecting the latest relevant laws takes way too long. Is it possible to automate legal review and document organization with AI?"
The company was handling dozens of overseas visa applications daily.
However, since all review processes were done manually, processing speed slowed down and small errors occurred frequently.
They had to individually verify whether applications had missing content, whether submitted evidence was sufficient, and whether everything met the latest immigration law standards.
These repetitive tasks consumed work hours and added stress to employees.
Most critically, small mistakes in this process could lead to visa rejection for clients.
Automation Scenario Built with Make + OpenAI
We built an automation using Make and OpenAI that handles everything from AI legal review of application drafts to reflecting the latest laws, evidence extraction, and revision suggestions in one stop.
(Since they had already built their website with Bubble, we integrated Bubble into the workflow as well.)
It receives requests from external systems, gathers application data and linked evidence files, and AI generates suggestions based on legal perspectives and review criteria.
All results are structured in the DB, ensuring reusability and tracking.
Make Scenario Details
Step 1. Loading Application Information and Immediate Response
Step 2. AI Legal Review and Suggestion Generation for Draft
The draft field is sent to OpenAI GPT CreateCompletion to review legal compliance and improvement points. 
AI returns the original text, revision reasons (in Korean), and revised version in JSON format. 
Bubble CreateDataThing saves the original, reason, and revised version to the suggestion record and links it to the application.
[Prompt]
You are a U.S. visa application expert with over 20 years of experience. Based on the content of the document below, please evaluate the following:
1. Legal compliance: Verify whether the information in the document complies with U.S. visa application standards and relevant laws, and identify any legally problematic areas or risk factors.
2. Strengthening approval likelihood: Identify elements in the document that could positively influence the visa review, and suggest any areas that could be supplemented or emphasized.
3. Evaluation criteria: All judgments must be limited to the content explicitly stated in the document.
4. Response format: Responses should be in
[{"original":"original content", "suggestion": "content and reason for proposed revision", "edit":"revised version of original reflecting the suggestion"}, {suggestion2}, {suggestion3}]
JSON format as an array.
5. The content in "edit" must be the revised version of the original. It should not simply be a suggestion for revision but the actual revised text reflecting the changes.
6. Language: All responses should be in English. Only the "suggestion" should be written in Korean.
Step 3. Evidence (Tab) Analysis and Correction/Supplementation Suggestions
You are a U.S. visa application expert with over 20 years of experience. Based on the content of the documents below, please evaluate the following:
- Error correction: Based on the tab document content, find any incorrectly stated or potentially misleading content in the draft and correct it properly.
- Content supplementation: Based on tab content, find areas in the draft where supplementary explanation is possible or better content can be written, and make suggestions.
- Evaluation criteria: All judgments must be limited to the content explicitly stated in the documents.
- Response format: Responses should be in
[{"original":"original content", "reason": "content and reason for proposed revision written in Korean", "revised":"revised version of original reflecting the reason"}, {suggestion2}, {suggestion3}]
JSON format as an array.
- The content in "revised" must be the revised version of the original. It should not simply be a suggestion for revision but the actual revised text reflecting the changes.
- If a specific part of the tab content was referenced, please also indicate in the reason which part was referenced.
- Language: Write "reason" in Korean. Write "original" and "revised" in English.
- Response: Ensure the response never gets cut off midway. Only respond in complete JSON format.
- If the investment amount is not clearly presented with specific figures, add detailed figures so that USCIS or the consulate can evaluate the substantiality of the investment scale.
- The definition of sufficient investment should be one year's operating costs needed to start a similar business in a similar area as the U.S. company.
- If actual operational evidence of the U.S. company is deemed insufficient, strengthen the explanation with evidence such as revenue data, employee count, and operational photos.
- Ensure the applicant's job description is specific enough by emphasizing concrete responsibilities and required skills to satisfy E-2 employee conditions.
- If the explanation of how the applicant's education and career connect to the U.S. position is insufficient, write suggestions to strengthen this.
Step 4. Status Update
Complete Flow at a Glance
- Automation starts when a request comes in.
- Application information is loaded and an immediate processing receipt confirmation is returned.
- AI performs a first-pass review of the draft from a legal compliance perspective.
- Evidence tabs are downloaded, text is extracted via OCR and organized for readability.
- The organized evidence and draft are sent together to AI to generate correction and supplementation suggestions.
- All suggestions are saved to the DB and linked to the application.
- Finally, the status is updated to 'Pending Review' to pass it to the next stage.
338 Hours Saved Annually, 92.9% Cost Reduction
Now the firm's staff no longer need to manually check dozens of documents by eye.
Simply uploading the application and supporting documents triggers Make's integration with OpenAI for automatic analysis, and the analyzed key data and AI feedback appear organized in Google Sheets in no time.
Staff just need to review the content and apply necessary revisions -- that's it.
Legal changes that were previously easy to miss are now automatically reflected by AI, eliminating the need to manually search for the latest standards.
Average review time per application reduced from 1-2 hours to 10-20 minutes!
Most importantly, employees are freed from repetitive, concentration-intensive simple tasks and can now focus on high-value work like client communication and strategy development.
Through this automation system, the company was able to save approximately 338+ hours of work time annually.
At an hourly rate of 20,000 KRW, this translates to approximately 6,760,000 KRW in labor cost savings.
Meanwhile, Make + OpenAI usage costs are approximately 480,000 KRW per year, resulting in over 92.9% cost efficiency.
Automation is the most reliable solution to break this vicious cycle.
This automation built with Make and OpenAI went beyond simply reducing time,
it created an environment where legal professionals can focus on what truly matters -- client response, strategy development, and problem solving.
An era where 'automation' is the standard for legal document work has arrived.
Stop worrying about whether you've missed something or made a mistake.
An automated system is already handling that for you.
IMPAKERS Blog | Slack Staff Management Automation Case Study Read More