What Do You Do When The Law Says You Can't Be An Entrepreneur?

International students on J-1 & F-1 visas face a serious limitation: they legally cannot run a business while studying in the United States. For students studying entrepreneurship, that’s like learning a language but not being allowed to speak it. They can learn how to build and launch a product, yet they cannot act on it when hands-on experience matters most.

At the same time, AI tools are reshaping how digital products get made and sold. Content creation, web development, and marketing are all shifting fast, and there is a real risk that international students will graduate having only read case studies about these tools instead of ever getting their hands on them. I didn’t want that to happen to my students, and here is how I legally overcome these problems.

Setting the Bar

My goal was specific. I set out to design a curriculum where students could build and market a real product from start to finish. They would create the product and reach paying customers. The curriculum had to meet all visa rules. It also had to work within one academic semester.

That meant solving a legal problem and an educational problem at the same time, and neither one could compromise the other.

The Idea: Publish Books, Not Businesses

Turning that idea into something legally sound took one more piece: I arranged for the university itself to hold the Amazon KDP (Kindle Direct Publishing) account, with all book revenue routed directly to the university rather than to any individual student. This was what made the whole model compliant by having students do the creative and marketing work, but never touch the money.

Talesha's ABC Book — **Talesha’s ABC book is the best-selling book of all the books created**

I picked children’s books specifically and paired the idea with AI tools. Writing and illustrating a children’s book with AI isn’t a trivial exercise. It forces real iteration and quality control on both text and images to reach something actually worth publishing. It gave students a concrete, finishable product, small enough to complete in a semester, but demanding enough to teach real AI-collaboration skills.

Building the Workaround

Turning that idea into something legally sound took one more piece. I arranged for the university itself to hold the Amazon Kindle Direct Publishing (KDP) account, with all book revenue routed directly to the university rather than to any individual student. This allowed the students to participate in every stage of a real product creation and sales operation without personally engaging in restricted commercial activity. It was the piece that made everything else possible.

**Though not technically required, having a university Amazon KDP account was pivotal in the school agreeing to the curriculum**

With that structure in place, I built out the rest of the pipeline as part of the curriculum. Students built their own websites and landing pages for their books. The university agreed to give the students a $100 ad spend budget, since all the revenues were going to the school. The students used this money to run paid Google Ads and paid social media campaigns to drive traffic. They also ran organic, non-paid social promotion alongside the paid efforts, so they’d understand both sides of customer acquisition, not just the one with a budget behind it.

The entire process had to take place within a single semester, including AI-assisted product creation, web development, and marketing. So the curriculum was built to move fast, with each phase feeding directly into the next: no book got marketed until it was published, and no landing page got built until the book existed to sell.

What Came Out of It

Over three semesters, students published 155 books on Amazon (see the students’ books at the end of this article). Each book was a real product supported by a marketing campaign that the students carried out.

Just as importantly, the legal structure worked without a hiccup. Across the life of the program, students participated fully in a genuine e-commerce operation, from product to storefront to paid ads, without a single visa violation. Routing revenue through the university from the start was what made this all possible.

Every student who went through the course left having personally built an AI-generated product from scratch, built the web presence to sell it, and run both paid and organic marketing campaigns to drive traffic to it. They learned how to do a basic but full-funnel skill set that students don’t usually do until after they graduate. All of this, and it was a lot, was done in a single term, despite a legal restriction that could have kept them on the sidelines.

A Critical Look at This Approach

No case study like this is complete without an honest look at where it struggled, and this approach to experiential learning had real friction points.

The first was scope. This was an intro-level class, and for many students, the semester’s first real task was simply learning how to use AI tools well enough to get usable output. That ate into time I’d planned to spend on marketing and execution, not tool literacy.

Timing was also an issue. This approach to my course happened at the very beginning of the generative AI wave, when AI-generated images were still easy to spot and genuinely hard to make look polished.

Some books came out looking rough. I made the call to let the students publish anyway. The point was for students to go through the full, real experience of shipping a product, warts and all, but it left a cleanup job behind. I’m now working with my TAs to go through the catalog, fixing up books that can be salvaged and pulling the ones that can’t.

Marketing became an area where I had to make some adjustments. In the first semester, I taught Google Ads for a week, which was not enough time for students to learn it well. By the final semester, I changed the approach. I explained why paid ads mattered and allowed students to choose how to spend their time and budget. Many focused on free social media efforts, which fit their constraints, but fewer students gained experience with paid ads.

The Amazon publishing process proved to be a recurring challenge. KDP’s submission flow required more support than expected. So I built a library of tutorial videos to guide students through formatting and submission. This was not something I had planned on, but it became necessary.

The hardest problem, though, wasn’t technical at all. These were entrepreneurship students, and what they actually wanted was to build their own businesses. Even after explaining exactly why that wasn’t legally possible for them, a number of students were not excited about the project simply because it wasn’t theirs in the way they wanted it to be.

Simplifying After a Year of Lessons

After a full academic year of running and adjusting this curriculum, I concluded the course needed to be simplified, not iterated on further. Trying to teach AI literacy, publishing logistics, web development, and digital marketing all in one semester was asking too much of an intro-level course.

My redesigned course splits the semester in two. In the first half, student teams sell a food product on campus, a fast, tangible way to learn how to create a simple product and get it to market quickly, without any of the platform friction that came with Amazon.

In the second half, teams move into a digital project, but instead of publishing books, they sell print-on-demand products through AmazeCommerce.com. This keeps the parts of the program that worked, including AI image generation and building a landing page. However, it removed the most time-consuming steps, allowing students to reach the selling phase faster.

The Takeaway

Looking back, the version of this course that created and shipped 155 books wasn’t the final answer. Instead, it was my first draft to solving a complex problem. And that’s really the point.

The visa restriction was never going to bend, so the curriculum had to. What started as a workaround for a legal constraint became, over a full year of trial and error, a lesson in something bigger: that the best solution to “you can’t do this the normal way” is rarely solved on the first try.

Selling children’s books proved that the core idea worked. Students could create a real product with AI, build the infrastructure to sell it, and run a real marketing campaign, all without ever touching restricted territory. Yet, proving an idea works and building the right version of it are two different things.

Building the right version of the course has required me to have some healthy humility. I had to recognize that an intro course could not carry the full weight of AI tools, publishing logistics, web development, and paid marketing within a single semester. The results looked strong in a list of 155 titles, but the pace placed too much strain on the learning process.

The new approach to the course is simpler and faster, and it works better for teaching. Students still work with AI, build a landing page, and take a product to market. They just get there without drowning in information.

There is a lesson here that extends beyond this course. A workaround that produces a strong outcome isn’t the same as a workaround that produces the strongest possible outcome for the students going through it. Knowing the difference and being willing to tear down something that worked to build something that works better is its own kind of entrepreneurship.

The Students’ Published Books

Please note that we are currently doing quality control on this long list of books. In the future, some of the titles below may no longer be on the list or lead you to a “no longer existing” Amazon page.