Law school class teaches students to build 6 legal AI prototypes with SIPOC and workflow gates

jkubicki · x · 2026-11-18

A law school instructor wraps up a fall class where six teams built six working legal AI prototypes with Ice Miller. His three-stage playbook: Thinking — decode the work first with SIPOC, recognizing every process has three versions (documented, described, actual); use Jobs to be Done beyond speed to cover lawyer identity and client perception. Planning — workflows must pass a gate (clear trigger, clear deliverable, repeatable, at least one judgment call and one handoff); map each step's inputs/outputs and internal customers; do build-or-buy checks where the case for building is usually personalization. Building — split instructions into separate files (audience, inputs, decision logic, outputs) for fixability and the 'defensibility and explainability' clients increasingly demand; distinguish instruments from agents.

Related event: Law School Course Ends with 6 Student-Built Legal AI Prototypes(2 posts)→

Original post →

More from coding & agent

coding & agent channel →