Law school AI course wraps: 6 student teams ship working legal AI prototypes

jkubicki · x · 2026-11-18

A law professor shares lessons from his fall course where six student teams built working AI prototypes with partners at Ice Miller. Key methodology: decode the work before touching a model (SIPOC mapping; the three versions of every process—written, described, actual); use Jobs-to-be-Done since value is more than speed; gate workflows on clear trigger, deliverable, repetition, at least one judgment call and one handoff; do a build-or-buy check where personalization (embedded lawyer judgment) usually justifies building; split instructions into separate files for fixability and the "defensibility and explainability" clients increasingly demand; and distinguish instruments from agents.

Original post →

More from coding & agent

coding & agent channel →