Ninety seconds
January 2024. A solo practitioner in Phoenix used an AI tool to review a 340-page commercial lease. Eleven clauses flagged, two contradictions identified, redline produced with UCC citations and three Ninth Circuit cases. Ninety seconds. Her previous method: eight to fourteen hours. She billed two hours verifying the output. Her old firm would have billed forty.
The billable hour was invented in the 1940s as a management tool for partners measuring associate productivity. It was never designed for clients. The entire edifice — $1,000/hour rates, 2,200-hour targets — is a wealth transfer mechanism running seventy years because bar associations control supply and define what constitutes practice. A paralegal with fifteen years cannot tell a client her landlord violated deposit law. A tool producing better analysis than a junior associate cannot be sold to consumers in most states.
AI does not fix the profession. It routes around it. A child can ask an LLM to explain a rental agreement more clearly than most legal aid offices can, given their caseloads. Eighty percent of civil problems are pattern matching. The facts vary; the framework does not. The restorative justice camp wants to change what happens in the room. We want to change whether you need the room.
Where we concede ground: The first wave of legal AI is reaching AmLaw 100, not the public. Efficiency gains accrue to partners, not people.
What would change our mind: A jurisdiction running AI-assisted civil services for five years with no improvement over current outcomes for unrepresented parties.
Read the full synthesis: What should law look like in 2050?