In part one of this series, I discussed AI assignments in business law, specifically one I integrated into my course called, “Written ApprAIsals.” When it came to this assignment, pioneering something groundbreaking wasn’t my goal. Instead, the motivation behind its development was to acknowledge reality. Since my students were already using AI, I wanted to teach them how to use it responsibly and transparently.
After running this assignment several times, I learned a lot about what worked, what didn’t, and what needed adjusting. The following is an update on the transformation of “Written ApprAIsals” into the new, streamlined assignment: “ApprAIsals.”
What stayed the same: The foundation
In both “Written ApprAIsals” and “ApprAIsals,” my students are assigned a pre-selected law. They must analyze its constitutionality while documenting their required AI usage.
While this foundation may have been solid, the resulting impact on my workload changed everything.
Where things broke down: The grading marathon I did not see coming
With “Written ApprAIsals,” my grading timeline looked roughly like this:
- Over 100+ students submitted their first drafts.
- Within two days, I returned detailed comments based on a multi-point rubric.
- Three days later, students submitted their AI logs based on initial attempts to improve drafts.
- Within two days, I provided comments on their AI logs with suggestions for improvement.
- Three days later, students submitted their final drafts.
- Within a few more days, I graded the final drafts using a similar rubric.
All of this happened within a very compressed two-week window.
I thought I was designing a thoughtful assignment for my students, only to find out I naively signed myself up for a grading marathon. This isn’t a criticism of my students, but of a flaw in my design. After several quarters of trying to make it work, it became clear “Written ApprAIsals” wasn’t viable.
The Adjustment: Same goals, sustainable structure
The biggest change I made was to stop asking my students to write the first draft. Instead, I now supply the first draft. I write an intentionally bad first draft, which goes against every instinct I have as a lawyer. Then, AI helps me create an even more flawed version. In other words, AI helps me produce exactly the kind of first draft that requires a careful, knowledgeable rewrite.
Once my students receive this flawed draft, they must:
- Identify all factual, legal, and analytical errors.
- Use their foundational knowledge of the law and AI to improve the first draft.
- Apply their understanding of the law.
- Write a polished, final draft free of errors.
Instead of a grading marathon, I can now evaluate their final drafts and AI logs together. This makes for a much more manageable and sustainable workload. The new structure acknowledges a simple concept: An instructor cannot meaningfully grade this much student output in three significant waves.
What didn’t change: The skills I want students to learn
While the structure of the assignment changed, my assessment of students’ understanding and application of the law remains the same. If a student’s final draft still contains errors from my intentionally flawed first draft, it tells me the student did not understand the legal concepts. My assessment of their foundational legal knowledge remains intact.
Additionally, I continue to teach my students to use AI responsibly. I know many students are feeding my entire first draft into AI. And I know AI sometimes validates my intentional mistakes, introduces new analytical errors, misses important exceptions, and even creates fictional legal precedents. Given these factors, my students must learn to recognize, correct, and verify the output they receive from AI. My assessment of their responsible use of AI stays in place.
My pedagogical goals stayed the same. And, the logistics finally do too.
An unexpected benefit: Students enjoy fixing my bad first drafts
Surprisingly, students like receiving the flawed first drafts. Some said it feels like they’re stepping into a supervisory role. Others said it’s more fun to “repair” something than to face the blank page. Many enjoyed spotting the mistakes they suspect AI made. My students’ engagement level didn’t drop. If anything, it improved.
Did I use AI in this article?
Let’s answer this question the same way my students would hopefully answer if someone asked them about “ApprAIsals.” Yes, AI was involved, but I — the human — did the work. In fact, writing this blog article mirrored the spirit of “ApprAIsals”:
- Start with something imperfect.
- Use AI as a tool, not a crutch.
- Revise, refine and fact-check.
- Produce a final version that reflects my tone and style.
And this is precisely what I want my students to learn.
Final thought: An adjustment that worked
I didn’t redesign “Written ApprAIsals” because it was pedagogically flawed. I redesigned it because it wasn’t logistically feasible for me to maintain. “ApprAIsals” keeps everything that matters to me:
- Teaching foundational legal concepts
- Training my students to engage in effective legal analysis and writing
- Emphasizing the responsible use of AI
Yet, it preserves these elements in a way that fits within the realities of teaching course sections. This wasn’t a bold leap forward. Rather, it was a practical, necessary step sideways. And it worked.
Written by Machiavelli (Max) Chao, Full-Time Senior Continuing Lecturer at the Paul Merage School of Business at the University of California, Irvine and Cengage Faculty Partner.