Get Your Hands Dirty First

TL;DR: Dan’s students grow more skeptical of AI every year. His response isn’t persuasion but sitting with and using the tools to consider their disciplinary implications.

What he’s doing: Dan Jackson teaches Artificial Intelligence for Lawyers: Uses, Risks, and Regulation, a roughly thirty-student lecture course built around hands-on practicums, which is now in its second run. He opens with a diagnostic: students mark their position about AI usage in general on a line from opponent to optimist, and nearly all of them land on the skeptic-to-opponent end.  Many have never opened an AI tool at all, a pattern he says has hardened rather than softened since 2022. 

His first practicum answers that directly. Each student builds their own custom GPT or Claude Project on any topic, explicitly not required to be law-related, then records a three-minute screen-share walkthrough; the strongest are presented in class. The point is less the artifact than the contact: students learn what the tool can do and where it breaks. The results ranged widely: a bot matching college athletes to name-and-likeness deals, pantry-based recipe and race-coaching tools, and a standout case-briefing bot one student trained on eighteen months of his own briefs so it would work in his personal style. For Dan, building the thing is what earns a student the standing to argue about it. “If you are concerned about what generative AI is doing to the legal profession,” he tells them, “you need to get your hands dirty with it and understand it.”

What’s working and what isn’t: The practicum reliably moves opponents past what Dan calls the skepticism hump, and free rein over the topic lets students find something that genuinely grabs them.  

He does something similar in his legal-design seminar,  where students build an AI “persona” to stress-test their prototypes. Students who compared AI-created “subject matter expert” bots against human experts found the bots about as useful, in some ways, as experts working under real time constraints. Though, students noted the bots tended to overproduce and needed reining in. 

Still, students have raised a significant tension. Why learn to do this if AI can do it faster? Dan focuses his answer on judgment, the kind that even the most technically proficient model still can’t replicate: “You need domain knowledge to refine the output, to ‘call BS’ or to make a brief ‘sing.’”

What’s next: Dan is keeping the build-your-own-bot practicum and adding one that uses a law-specific commercial tool to generate a case brief. He’s also weighing whether to make the open practicum subject-specific, trading the creative free rein for something more directly useful to legal training.

Have you tried something new with AI? How did it go? Send an email to [email protected] to let us know!