LawDroid Manifesto Podcast The Claude-Native Lawyer: Zack Shapiro
9 snips
Mar 9, 2026 Zack Shapiro, founder of Raines LLP and author of the 'Claude-Native Law Firm' piece, explains how he runs a two-person firm using agentic AI and custom skills. He discusses why prompting quality matters, practical Claude-native workflows for redlines and research, how AI reshapes junior roles and firm structure, and concrete steps to build and iterate AI skills for daily legal work.
AI Snips
Chapters
Transcript
Episode notes
Input Quality Beats Model Training
- The secret to better AI legal output is superior input not bespoke training data.
- Zack Shapiro explains that lawyers' skill at precise instructions (prompts) produces far better results than training models on firm documents.
Fictional M&A Example Generated With Claude
- Zack fictionalized an M&A example with Claude to illustrate workflows he couldn't share directly.
- He used Claude to create a safe, representative anecdote that preserved client confidentiality while showing capabilities.
Iterate Skills Like Reinforcement Learning
- Build reusable AI skills by iterating: teach Claude your preferences, test outputs, then update the skill when it misbehaves.
- Zack says he treats skills like reinforcement learning where defaults compound value across tasks and time.