Episode 017 Training AI to Think Like Your Practice

What are you actually doing with AI — and does it work?

Ron and Heather put their tools down long enough to talk about what they've been building with them. This isn't a roundup of tools you might try. It's an inside look at two practitioners who have spent serious hours creating AI-powered workflows for their own legal and legal-adjacent businesses — and what those projects revealed about where AI is actually useful for law firms right now.

In this episode:

• Heather's report from the Maryland Legal Summit — how attorney AI adoption shifted dramatically in a single year, and what the fear conversation looks like now

• Why hallucination is more than a citation problem — the 21 ways AI can corrupt a legal brief, from wrong standards of proof to mutated judicial language

• Heather's 11-module Bankruptcy Paralegal Course, built with AI, including a Claude Code-built floating chatbot trained on the course content

• How Heather's paralegal team uses Gemini inside Google Sheets to auto-generate and schedule weekly client status reports — eliminating a manual step entirely

• Ron's markdown workflow system for podcast post-production — built in Claude, portable to any AI environment

• Ron's Case Assessment Pack: a deep-dive workflow that turns a client intake recording into a preliminary liquidation analysis, exemption review, and client-ready deliverable

• This week's Practice Signal — a 9th-year BigLaw associate terrified about making partner and how AI could be the rainmaking engine he hasn't considered

• The FSJ breakdown for capturing firm knowledge — what Flintstones, Simpsons, and Jetsons lawyers each need to do right now to prepare for the agentic AI era

We also discuss:

• OWLL, the recording app Ron just downloaded — and why he thinks lawyers should be recording everything

• Otter.ai as the baseline for meeting intelligence, and how Ron's workflow pack goes further

• The difference between using AI as a Google machine versus as a strategic collaborator

• Why opposing counsel — not the filing attorney — is catching most hallucinated citations right now

• Harvey and Legora showing up at the Maryland Legal Summit — and who's actually using them

Key Takeaway

AI doesn't magically understand your practice. It inherits whatever you've ingested into it. That's the core lesson from everything Ron and Heather describe — the tools work because they were loaded with domain knowledge, firm context, and real workflow logic. Without that, you get a very confident machine that doesn't know what it doesn't know.

For Flintstones lawyers, the move is simple: start documenting. Write down what you do. Create checklists. Capture the firm knowledge that currently lives in your head. Simpsons lawyers need to organize that knowledge — standardize file names, define workflows, build taxonomies. Jetsons lawyers are ready to connect the systems, build repositories, and create governance. The future belongs to firms that treat data as infrastructure. Where you start matters less than whether you start.

Mentioned in This Episode:

• Maryland Legal Summit / Maryland State Bar Association

• Claude (Anthropic)

• Claude Code

• ChatGPT (OpenAI)

• Gemini (Google) — including Google Sheets integration

• Microsoft Copilot

• Harvey

• Legora

• Otter.ai

• OWLL (recording app)

• BusinessGPT

• BenchSimAI (Chris Ryan) 

• Foundation AI

• Reddit (Practice Signal source)

• Cromwell case — hallucinated citation example 

• Heather's Bankruptcy Paralegal Course (Propel AI)

• Ron's AI Builds: Ground Zero episode with the Free Motion to Extend Workflow

• Google Workspace (Sheets, Drive, Docs)

[email protected]