Advertise on podcast: AI Tools for Practicing Lawyers

SHOW NOTES Every legal tech conference wants to talk about AI drafting and contract review. Almost nobody talks about the business behind the practice: the virtual assistant in the Philippines, the outsourced paralegal, the 1099 you found through a Facebook ad. Natalie Dalton, founder of Out of the Box HR, says that's the blind spot. Lawyers keep buying software before they understand how their own firm actually works. Her rule is simple: diagnose before you digitize. So what happens when you plug AI into a process that was broken to begin with? In this episode: - Why HR is really organizational design, not just payroll, benefits, and handbooks - Whether a solo lawyer with an outsourced paralegal and an offshore assistant needs HR at all - Employer of record arrangements and the compliance risk of hiring overseas workers directly - Starting with strategic objectives: defining roles, decision rights, and accountability before you hire - Using ChatGPT and Claude as a planning partner, and why an AI-drafted employee handbook only gets you about 80% of the way there - AI resume screening: setting must-haves and deal breakers, and keeping human eyes on the shortlist - Natalie's time at Document Crunch, an early AI platform that put construction contract knowledge in the hands of field teams - Natalie's AI Decision Tree and Law Firm AI Playbook, built around "diagnose before you digitize" - Reactive vs. proactive firm owners, and the hidden time bomb of a disengaged employee - Using AI to structure one-on-ones and turn meeting notes into follow-up action items We also discuss: - Why an AI that always tells you you're right is a risky HR advisor, and how Heather's AI assistant Bosley pushes back - AI memory failures, even with instructions stored in GitHub - Adding an HR consultant to the classic startup team of insurance agent, lawyer, and accountant - How Heather tested AI resume screening against a human assistant before trusting it - The 2026 midterm ballot measures on AI data centers, and whether slowing AI development would be such a bad thing - AI as "durable infrastructure," and why "AI app" will soon sound as dated as "internet app" - Voters in 13 states weighed 50 local ballot measures on AI data centers in 2026, and nearly every decided result went against development. Is slowing AI down such a bad thing? Ballotpedia — Data center-related ballot measures, 2026 Download: Natalie Dalton's AI Decision Tree and Law Firm AI Playbook A practical framework for deciding where AI belongs in your firm before you spend money on software. - Starts with the business, not the tool - Identifies where time and expertise are being wasted - Separates technology problems from management, training, and workflow problems - Helps you avoid automating a process that doesn't work Key Takeaway Natalie's point is blunt: install technology on top of an ineffective process and you just make that process ineffective faster. Before you buy another subscription, figure out whether your real problem is the tool, an unclear role, a broken workflow, or a person nobody is managing. Mentioned in This Episode - Natalie Dalton, founder of Out of the Box HR - Natalie's AI Decision Tree and Law Firm AI Playbook - Document Crunch - ChatGPT - Claude - Anthropic and OpenAI - Propel (outsourced paralegal services) - Indeed - Bosley (Heather's AI assistant) - Hugging Face (discussed in [Episode 033]) - 2026 midterm ballot measures on AI data centers - Jen Grondahl Lee (returning guest) Highlight Reels on YouTube and Instagram [email protected]

SHOW NOTES Students in 2026 can't imagine writing a ten-page paper without ChatGPT. Judges are split between refusing to touch AI and quietly using it to draft. Lawyers keep filing hallucinated citations despite years of headlines and sanctions. Maura Grossman, the lawyer and computer scientist who helped prove machine learning could outperform human document review, joins Ron and Heather to separate what AI does well from what has to stay human. Is human judgment actually wiser than AI judgment, or is it just a bigger pile of data? In this episode: - The state of the bench in October 2026: a "wild west" of judges who refuse AI, judges who use it freely, and almost no uniform guidance - Why judicial resistance to AI is mostly fear, and how an hour of hands-on training changes it - Heather's challenge: lawyers miscited cases for generations, so shouldn't the standard focus on the final work product rather than how it was produced? - Whether pro se litigants deserve more slack than licensed attorneys for hallucinated citations - Why harsh sanctions haven't reduced hallucinations, and a judge who solved one with a simple warning - The "sandwich approach": the judge thinks first, AI helps in the middle, the judge decides at the end - The live debate over whether prompting is dead, and why "just talking to your AI" is still prompting - Maura and Gordon Cormack's research showing generative AI's claimed edge over older machine learning in e-discovery largely disappears in an apples-to-apples comparison - Why human judgment matters: alignment, culture, morals, and Bostrom's paperclip problem - What happens to mentorship and learning when AI does the research a junior lawyer used to do We also discuss: - A law student who admitted he never read a single case in three years, summarizing everything through ChatGPT - Maura's classroom fix: a personal journal where AI use is prohibited - Clients who arrive convinced by ChatGPT that their case is worth a million dollars - The lawsuit alleging ChatGPT told a litigant she was being snowed by her lawyer - Whether AI companies should compensate authors and artists whose work trained their models - Practice Signals: a new solo litigator who wants to add contract drafting with zero transactional experience Key Takeaway Maura's closing advice cuts through most of the noise: the problem should drive the solution. Don't buy AI because the firm down the street has it. Find the pain point in your own practice, then find the tool that fixes it. And whatever tool you pick, it can't do the thinking a judge or lawyer is paid to do. That stays with you, at the start and at the end. This season the show shifted from Flintstones, Simpsons, and Jetsons to users and builders, and this episode lands on both. Users need to know enough to verify what the tool hands them. Builders need to remember that a beautifully engineered workflow doesn't make them competent in an area of law they've never practiced. Jane Jetson may think prompts are so March 2026. Maura's research says the jury is still out. Mentioned in This Episode - Maura R. Grossman, Research Professor, School of Computer Science, University of Waterloo; Adjunct Professor, Osgoode Hall Law School; Visiting Professor, Duke Law School - The Judicious Judge's Guide to Generative Artificial Intelligence and Large Language Models (forthcoming) - "Utopia or Dystopia" paper on self-represented litigants and AI - Grossman and Grimm general-principles paper, Sedona Conference Journal (2025) - Grossman and Cormack, 2011 study on technology-assisted review in e-discovery - Judge Paul Grimm (Podcast Episode 023 Deepfakes and the Liar's Dividend) - Gordon Cormack, University of Waterloo - Judges Maritza Braswell (Colorado), Scott Schlegel (Louisiana), and Xavier Rodriguez (Texas) and their judges-only AI group - Federal Judicial Center AI toolkit for federal judges - National Center for State Courts and Thomson Reuters Institute - Georgia Bench and Bar Conference - MIT research on brain engagement when writing with LLMs - Nick Bostrom's paperclip maximizer thought experiment - Federal Rule of Civil Procedure 11 - Nippon Life lawsuit against OpenAI - ChatGPT, Claude, Gemini, Vertex AI, Google Cloud, Microsoft - Harvey, Legora - Westlaw, Lexis - Reddit  Highlight Reels on YouTube and Instagram [email protected]

SHOW NOTES Every AI tool lawyers use was trained on the whole internet, including the good sites, the bad ones, and the ones that say the moon landing was faked. Jonathan Schaeffer has spent more than 40 years in AI and helped teach computers to master checkers and poker, and he thinks that is exactly the problem. His answer is Kind, an AI that runs entirely on your own desktop, never touches the cloud, and knows nothing except what you give it. Ron and Heather press him on whether a walled-off AI really solves the problems lawyers face. Then the conversation turns to AI agents that did not stay walled off at all. Is the safest AI for a law office the one that knows the least about the outside world? In this episode: • Why Jonathan treats privacy, safety, security, and accuracy as non-negotiable requirements for any AI tool • How Kind works: an AI that runs on your desktop, answers only from your own files, and never sends data to the cloud • The difference between Kind Hybrid and Kind Local, and what fills Kind Local's six-gigabyte footprint • Jonathan's live demo for an Alberta bar group, where an ethics question about Alberta came back with Maryland citations • Jonathan's claim that source-linked answers drawn from a closed collection can't produce a hallucinated citation, and Ron's pushback on whether that is even the right use case • Ron and Heather challenge Kind's auto-updating legal collections: does linking to a live website break the walled garden? • The consistency checker, which uses an idle desktop to find logic and fact contradictions across a collection of files • Law firm use cases: client files that never leave the machine, and matter-specific collections paired with one jurisdiction's law • The open letter on AI cybersecurity risks Jonathan signed, and Ron's concern that calls for scientific unity can be used to silence dissent • Jonathan's sobering prediction that it may take a disaster before governments regulate AI We also discuss: • The July 2026 Hugging Face incident, in which OpenAI test agents escaped a contained environment and broke into outside systems • Why Jonathan compares escaped AI agents to a virus that mutates in minutes rather than months, and why a kill switch is not a real answer • Practice Signals: a lawyer on Reddit asks how to show a 16-year-old in another state what practicing law is really like • Ron's method: feed Delaware's bar admission clerkship checklist into an AI and have it build a version for a high school student • Heather's pitch for Chris Ryan to spin BenchSim into a resource for aspiring lawyers • Jonathan's closing case for keeping the human in the loop, especially in education Key Takeaway Jonathan's strongest point is about authority, not privacy. An AI trained on the whole internet will answer an Alberta ethics question with Maryland law and sound confident doing it. Availability is not authority. A tool that can only answer from the files you load forces you to decide what the law is before you ask the question. As Ron pointed out, that is what lawyers are supposed to be doing anyway. That opens a door for Flintstones and Simpsons lawyers who have stayed away from AI because they won't upload client files to someone else's server. The catch is that the walled garden only works while the walls stay up. Once you link a live website or connect collections, you are making the same judgment call you would make with any cloud tool. Jetsons lawyers will still want the reach of internet-connected tools. Everyone else has to decide which work belongs inside the wall. Mentioned in This Episode • Jonathan Schaeffer, Professor Emeritus, University of Alberta; Founder and CEO, Synsira Software Solutions • Kind by Synsira (Kind Hybrid, Kind Local, and the forthcoming Kind Shared) • Alberta Machine Intelligence Institute (Amii) • ChatGPT, Claude, and Gemini • OpenAI and Meta • The Hugging Face incident (July 2026) • METR's independent investigation of the OpenAI/Hugging Face incident • Instadocs: AI Gone Wild (Netflix, 2026), a 27-minute documentary on the Hugging Face hack • The AI Doc: Or How I Became an Apocaloptimist (documentary) • Open letter from scientists on AI cybersecurity risks • Chris Ryan and BenchSim (featured in Episode 016: Can an Al Judge Train You Better Than the Real One?) • Delaware bar admission clerkship checklist • Reddit (source of this episode's Practice Signals story) [email protected] Highlight Reels on YouTube and Instagram

SHOW NOTES Google or Microsoft? Most lawyers treat that as a question about Outlook versus Gmail. It isn't. It's a question about who protects your clients' confidential information, and the honest answer surprises people: neither platform wins on paper. The real question is whether you bought the right protections, turned them on, and trained your people to use them. In this episode: - Why passkeys and phishing-resistant authentication belong near the top of a law firm's security shopping list - How one "anyone with the link" share can expose a client's tax returns, bank statements, and bankruptcy schedules - Data loss prevention (DLP): what it does, and why both Google and Microsoft tie it to specific subscription tiers - The lost-laptop scenario, and why Microsoft 365 Business Premium (Intune plus Defender for Business) gives Microsoft an edge for Windows-based firms - Third-party AI apps that ask for access to your Google Drive or OneDrive, and who in your firm has authority to approve them - Gemini and Copilot: why AI assistants inherit your existing permission problems and make them easier to exploit - Why the cheapest subscription may not be cheaper in any meaningful sense - Five questions every firm should be able to answer about its security posture We also discuss: - "Folder mania" and the stakes of giving an AI app access to entire folders - Microsoft Purview sensitivity labels for classifying and protecting confidential documents - Why Ron, a longtime Mac user, isn't sure Microsoft's device-management advantage applies to him - Reviewing contractual commitments on data use, retention, and model training before enabling enterprise AI - The three-legged stool of AI security: technology, configuration, and people Companion Resources: AI Builds on Google Workspace and Microsoft 365 Ron's companion AI Builds episodes walk through the specific settings behind this comparison, with materials to help you find them. - AI Builds: How Secure Is Google? - AI Builds: How Secure Is Microsoft? - Companion materials Google - Companion materials Microsoft Key Takeaway Security isn't a brand you buy. Google and Microsoft both offer excellent technology. Whether it protects your clients depends on which edition you paid for, whether anyone configured it, and whether your people follow the rules. A locked front door does nothing if your paralegal hands out the key in a shareable link. Flintstones lawyers who assume their email provider handles security by default need to look again. Simpsons lawyers about to switch on Gemini or Copilot need a permissions review first, because the AI will find every overshared folder. Jetsons lawyers connecting third-party apps to their files need to know exactly which ones have access and who approved them. Mentioned in This Episode - Google Workspace - Google Drive - Gmail - Gemini - Microsoft 365 - Microsoft 365 Business Premium - Microsoft OneDrive - Microsoft SharePoint - Microsoft Outlook and Word - Microsoft Copilot - Microsoft Purview - Microsoft Intune - Microsoft Defender for Business - Microsoft Entra - AI Builds companion episodes on Google Workspace and Microsoft 365 (linked above) - The three-legged stool of AI security [email protected] Highlight Reels on YouTube and Instagram

SHOW NOTES Your email is in Outlook. Your briefs are in Word. Your client files sit in SharePoint and OneDrive. Now Microsoft wants to put Copilot in the middle of all of it, and Copilot will find anything your permissions already let you see. In this AI Builds episode, Ron Drescher argues that the real risk isn't Microsoft's security. It's whether your firm ever configured it. How secure is Microsoft, and how secure is your firm's use of it? In this episode: - The three-legged stool applied to Microsoft 365: Microsoft secures the platform, your firm configures it, and every user still holds a key - What Microsoft 365 actually protects (encryption, multifactor authentication, sign-in monitoring, auditing) and why your subscription tier decides which protections you have - The two questions to ask about any Microsoft security feature: does my plan include it, and is it turned on? - "Folder mania" in the cloud: how years of loose SharePoint and OneDrive permissions become a live problem once AI can search them - Why Copilot can surface a confidential client file without breaking a single rule: it followed the rules, but the rules were wrong - Sorting out the Copilot family: consumer Copilot, Microsoft 365 Copilot Chat, and the fully integrated Microsoft 365 Copilot - What you are really authorizing when you click "Connect your Microsoft 365 account" on a third-party legal AI tool - User-level vs. application-level permissions, and four questions to ask any AI vendor before granting access - The five admin locations a small firm should know: Microsoft 365 Admin Center, Microsoft Entra, SharePoint Admin Center, Microsoft Defender, and Microsoft Purview We also discuss: - Why cloud storage is not a backup or disaster recovery plan - Ransomware as a business continuity question: could you meet tomorrow's filing deadline? - Why you shouldn't change unfamiliar security settings in the middle of a workday - The "miscellaneous" folder nobody has opened since 2017 - Five questions to bring to whoever manages your firm's technology - The AI paradox: new tools make old security mistakes more expensive Download: How Secure Is Microsoft? A Microsoft 365 Security Guide for Solo and Small-Firm Lawyers A nine-page companion to this episode (and to How Secure Is Google?) It walks you and your IT provider through six places to review in Microsoft 365, with Microsoft-published screenshots and exact menu paths. - Users, administrators, and stronger sign-in through security defaults or Conditional Access - Client sharing limits in SharePoint and OneDrive, and the permissions you've already granted to outside apps - Oversharing reports to run before Copilot, plus how to confirm Purview auditing actually works - A plain-English licensing page and a printable firm security checklist https://lawyeraitoolkit.com/deliverables#row-WoaNrpcH4Y Key Takeaway Copilot doesn't create your permission problems. It exposes them. A SharePoint site that has been quietly overshared for five years was a low-grade risk when nobody could search it. Put an AI assistant on top, and an associate can surface another client's confidential file by asking one question. Microsoft built the lock. Your firm decides who holds the keys. Simpsons lawyers feel this first. They've turned on Copilot or connected a new legal AI app without ever opening Entra or the SharePoint Admin Center. Flintstones lawyers running Microsoft 365 out of the box carry the same oversharing risk; they just haven't triggered it yet. Jetsons lawyers building on Microsoft data need to know exactly what permissions their tools hold. The answer is yes to AI inside Microsoft 365, if you've audited who can open the folders first. Mentioned in This Episode - Microsoft 365 (Business Basic, Business Standard, Business Premium) - Outlook, Word, Excel, PowerPoint, Teams - Exchange Online, OneDrive, SharePoint - Microsoft Entra - Microsoft Copilot (consumer), Microsoft 365 Copilot Chat, Microsoft 365 Copilot - Microsoft 365 Admin Center - SharePoint Admin Center - Microsoft Defender - Microsoft Purview - Companion episode: AI Builds: How Secure Is Google? - Prior episode on "folder mania" - Microsoft's security documentation Microsoft 365 Copilot: Data, Privacy, and Security — Microsoft's explanation of how Copilot handles organizational information, protects prompts and responses, and respects existing Microsoft 365 permissions. Particularly relevant to the distinction between using AI and giving AI access to your firm's documents. Microsoft: Enterprise Data Protection for Copilot — Explains Microsoft's enterprise data protection commitments, including restrictions on using organizational prompts and responses to train foundation models. Useful for distinguishing business Copilot experiences from consumer AI services. Microsoft Entra: Permissions and Consent — Explains how third-party applications obtain permission to access Microsoft 365 information, including the important distinction between delegated permissions and application permissions. This is especially relevant when a lawyer connects an outside legal AI product to Outlook, OneDrive, or SharePoint. Microsoft Cloud: Shared Responsibility Model — Explains which security responsibilities belong to Microsoft and which remain with the law firm, including user identities, device security, access permissions, and configuration. This is the official Microsoft resource supporting our Three-Legged Stool concept. Microsoft Learn Microsoft: Manage Copilot Chat — Explains how administrators can manage employee access to Copilot Chat, including restrictions and application availability. Useful for lawyers who want to control when and how AI becomes available within their Microsoft environment. Microsoft Learn Microsoft: Copilot Security and Governance — Explains how Microsoft addresses excessive document sharing, AI access to organizational information, and compliance controls. Particularly relevant to our Foldermania discussion and the risk of Copilot discovering documents that employees already have permission to access but shouldn't.  [email protected] Highlight Reels on YouTube and Instagram

SHOW NOTES Your clients' bank statements, tax returns, and credit reports may already live on Google's servers. Gmail, Drive, Docs, and now Gemini. Google probably protects that data better than the server in your closet ever could. But Google securing its infrastructure doesn't mean your client's information is secure. **If Google is doing its job perfectly, who's responsible when client data still gets out?** **In this episode:** - The four separate security worries lawyers lump together: outside hackers, Google employees, AI model training, and your own firm handing out access - Why Google's infrastructure likely beats the typical small-firm setup of a closet server, an on-call IT consultant, and an untested backup - Google's shared responsibility model: Google secures the platform, and the firm controls who and what gets in - How a mis-shared folder, a compromised account, or an over-permissioned app exposes client data while Google's systems work perfectly - Limited-scope authorization and the drive.file permission, which give an AI app one matter's files instead of the whole filing cabinet - Three Gemini environments with different data terms: consumer Gemini, Gemini under Workspace business terms, and Vertex AI - Why "doesn't train on your data" is not the same as "doesn't retain your data," including abuse monitoring, caching, and 30-day grounding retention - What a security readiness review of Ron's own AI bankruptcy platform revealed, and why the answer to "ready for real client data?" was no - Four failures Google's infrastructure can't fix: one firm seeing another firm's files, a departed paralegal who still has access, temporary copies never deleted, and no incident response plan - Five things lawyers should do now: know your account, require strong sign-in, check third-party permissions, question AI vendors, and treat security as ongoing **We also discuss:** - Why having a Gmail address doesn't put you under Workspace business protections - Consumer Gemini configurations where humans may review your conversations - Why zero data retention on Vertex AI takes affirmative configuration choices - Passkeys and security keys as phishing-resistant sign-in, even though multi-factor authentication is a pain - The questions every AI vendor should be ready to answer, including what happens to your documents when you cancel > **Download: AI Builds — How Secure Is Google? (Three-Legged Stool Visual Guide)** > A six-page, click-by-click guide to Google's admin console for locking down client data, organized around Ron's three-legged stool: Google's security, the application's security, and the firm's own practices. > - Requiring 2-Step Verification and passkeys across the firm > - Controlling who in the firm can use the Gemini app > - Reviewing which third-party apps hold OAuth access to firm data, and how much > - Setting external Drive sharing boundaries and Gemini conversation retention **Key Takeaway** Google can protect the vault. It can't stop you from handing someone the key. Ron's own platform had good safeguards on paper. It had limited Drive access, firm isolation, and an audit trail, and it still wasn't ready for real client data, because designing protections and proving they work are two different things. The question isn't whether Google is secure. It's whether every link in the chain touching your client's information is secure. For Users, the work is unglamorous. Know which account you're signed into, turn on strong sign-in, and read what an app is asking for before you click Allow. For Builders, the bar is higher: test firm isolation, offboarding, deletion, and incident response before the first real client file arrives. And expect lawyers to start asking how you tested those protections. **Mentioned in This Episode** - Google Workspace (Gmail, Google Drive, Google Docs) - Gemini (consumer) - Gemini in Google Workspace business accounts - Google Vertex AI - Google's shared responsibility model (Google Cloud documentation) - Google's drive.file limited-scope permission - Google Workspace privacy documentation and Vertex AI enterprise terms - Passkeys and security keys - Ron's AI-assisted consumer bankruptcy platform and its internal security readiness review Google Workspace Generative AI Privacy Hub — Google's commitments concerning training, confidentiality, human review, and Workspace data. Google Cloud: Vertex AI and Zero Data Retention — Training restrictions, caching, retention exceptions, and configuration requirements. Google: Choose Google Drive API Scopes — Explains how developers can limit an application's access to individual files. Google Cloud Shared Responsibility — Explains what Google secures and what customers remain responsible for. Google: Turn the Gemini App On or Off — Explains the differences in protection between qualifying Workspace accounts and other configurations. Highlight Reels on YouTube and Instagram > [email protected]

SHOW NOTES Every vendor at next year's bankruptcy conventions will say their product "uses AI." The better question is what the AI is actually allowed to do without a lawyer signing off. Colorado consumer bankruptcy attorney Matt McCune, a returning guest, gives a live first look at BK Fastlane, the AI-driven intake and document platform he built for consumer bankruptcy practices. Throughout, Ron keeps returning to one question: where does the lawyer stay in control? In this episode: - The debtor-facing secure portal: account creation, multifactor authentication, firm branding, and "Sarah," the AI onboarding agent - An intake form mapped to the schedules, the voluntary petition, and the Statement of Financial Affairs, with required fields and hard stops - A "never ask for a document you don't need" rule: no mortgage statement if there's no house, no pay stubs if there's no job - How the client screws up the form, and how the software gets them back on track - The attorney-facing CRM workspace and its four-stage pipeline, from new lead to ready for petition prep - AI document review: the AI suggests accept or reject, but a human approves every document, including high-confidence ones, "for now" - The automated chase list: daily follow-up emails for rejected or missing documents, so nobody on staff has to call the client - The AI tab: an instant means test analysis, a confidence score, and a case complexity score - Matt's cardinal rule: the AI never takes an action, especially with the client, without lawyer approval - Ron's question from his Infusionsoft days: how far do you let rules-based client emails run on their own once AI can customize them? We also discuss: - Two intake paths: a free consult for routine cases, and a paid $250 consultation for complex or messy ones - The three-phase roadmap: intake and CRM now, post-filing workflows next, electronic filing and petition prep last - Why a bankruptcy-only "vertical" product can do things a horizontal platform like Clio can't - Passkeys vs. multifactor authentication, and the bankruptcy court's new authenticator app requirement - Users vs. Builders: why Matt wants early-adopter "builder" firms now and Flintstones firms after all three phases ship - Matt's years of suing bad-actor bankruptcy lawyers under Sections 526, 527, and 528, and why that shaped his mission for the product Watch the full demo on YouTube: https://youtu.be/pEbOWwf8tHM Click here for a PDF slideshow of the software demo Key Takeaway The most important thing in this demo isn't a feature. It's the approval button. Matt built BK Fastlane so the AI can read a pay stub, flag a fuzzy Social Security card, and run a means test, and still not move a document or send an email until a human says yes. His answer to whether high-confidence documents should skip review was blunt: not yet. "I don't trust it. I don't think anyone should trust it." Matt himself drew the line between Users and Builders. Right now he wants early-adopter firms that are curious about how this gets built but have no time to build it themselves. His target for Flintstones lawyers is roughly a year out, once all three phases ship. If you're a Simpsons-stage consumer bankruptcy lawyer still stitching together four or five products, this is a look at where single-practice-area software is heading. It is not a recommendation to switch today. Mentioned in This Episode - Matt McCune, consumer bankruptcy attorney (Colorado), founder of BK FastPass - BK Fastlane, the first product from BK FastPass - NACBA (National Association of Consumer Bankruptcy Attorneys) and its April 2027 convention in Orlando - Microsoft Azure - Claude Cowork - Clio, Clio Grow, Clio Manage, Clio Jubilee - Lawmatics - NextChapter - Best Case - PracticePanther - Rocket Lawyer - Infusionsoft (now Keap) - Zillow - BAPCPA (2005) - 11 U.S.C. Sections 526, 527, and 528 [email protected] Highlight reels: YouTube: https://www.youtube.com/@AIToolsforLawyers Instagram: https://www.instagram.com/aitoolsforlawyers

SHOW NOTES Maryland has about 40,000 lawyers. Roughly half a percent of them work in civil legal aid. Add more lawyers and you still hit a ceiling, because every lawyer can only serve so many people. Reena Shah, Executive Director of the Maryland Access to Justice Commission, joins Ron and Heather to ask what happens when technology breaks through that ceiling, and what happens when the general-purpose chatbots break through it first. What if AI's biggest contribution to law isn't making lawyers faster, but reaching the people who never had a lawyer at all? In this episode: - The civil justice gap in Maryland: no right to counsel in civil cases, 40% of households unable to cover basic needs, and legal aid organizations meeting roughly 20% of demand - Maryland's AI and Access to Justice Summit, one of the first state commission events of its kind, and why it was run as a hands-on workshop instead of a lecture - What a pre-summit survey of 187 legal aid staff revealed: few AI policies, few enterprise tools, and no clear organizational stance on staff use - Why Maryland Legal Aid's Vicki Schultz says policy comes first, and why that policy has to start with the organization's values - Self-represented litigants walking into Court Help Centers with 20-page AI-generated pleadings, and refusing to believe the lawyers who tell them it's wrong - The Innovations in Tiered Legal Services Task Force and its push for a UPL waiver to test safe, closed AI tools in low-risk areas of law - Ron and Heather's disagreement: teach pro se litigants to use AI responsibly, or hand them a vetted tool? And is the teaching itself UPL? - The Nippon Life lawsuit against OpenAI, and why Reena argues the big AI companies crossed the UPL line without ever asking permission - Users vs. builders for legal aid: Maryland Legal Aid's home-grown MLA GPT, rent ledger reviewer, and brief creator versus off-the-shelf tools like NotebookLM - A TurboTax-style expungement tool as a model for scaling high-volume legal help We also discuss: - Anthropic's Claude for Legal launch and the connectors and skills it included for pro bono and legal aid organizations - Whether AI companies should give legal aid organizations pro bono or reduced-cost access, and who is actually asking them - Why AI is a prediction model, not a search engine, and how that distinction shapes training - The People's Law Library and the work of making legal websites accurate enough for AI tools to pull from - Practice Signals: a Reddit lawyer asks where the access-to-justice gaps are, and Ron adds a fourth question about what AI can fill - The AI Doc: Or How I Became an Apocaloptimist, and a preview of the show's 50th episode Key Takeaway Reena's argument is about math, not efficiency. Every lawyer has an upper limit. Technology is the only thing on the table that scales past it, but only if it's built safely and vetted before anyone relies on it. Meanwhile the general-purpose chatbots are already giving legal advice to people who can't tell a good answer from a confident one. Availability is not authority. The pro se litigant arguing with the Court Help Center lawyer hasn't learned that yet. For users, the lawyers running off-the-shelf tools, this episode is a warning about who is sitting across the table: more and more often, it's someone holding an AI-drafted answer they trust more than you. For builders, Maryland Legal Aid shows what a small team with real subject-matter expertise can do at low cost: narrow tools aimed at narrow, repetitive problems. Solo and small firm lawyers live with the same constraint legal aid does. There's more demand than there are hours. Mentioned in This Episode - Reena Shah, Executive Director, Maryland Access to Justice Commission - Heather Gardner, co-host and summit speaker - Maryland AI and Access to Justice Summit - Innovations in Tiered Legal Services Task Force (Maryland Access to Justice Commission, Maryland Judiciary, Maryland State Bar Association) - Maryland State Bar Association - Maryland Legal Aid and MLA GPT - Vicki Schultz, Maryland Legal Aid - John Jeffcott, Maryland Legal Aid - Donald Tobin, former Dean, University of Maryland School of Law - Chief Justice Matthew Fader, Supreme Court of Maryland - Lindsay Bramble - Angela Tripp, National Center for State Courts (formerly Legal Services Corporation) - Legal Services Corporation - Stanford University national summit on AI and access to justice - Maryland Volunteer Lawyers Service (MVLS) - Maryland Court Help Centers and county family self-help clinics - People's Law Library of Maryland - Anthropic and Claude for Legal - OpenAI and ChatGPT - Google Gemini and NotebookLM - Grok - Nippon Life Insurance Company of America v. OpenAI (N.D. Ill.) - The AI Doc: Or How I Became an Apocaloptimist (2026) - Reddit (source of this episode's Practice Signals post) Highlight Reels on YouTube and Instagram YouTube: https://www.youtube.com/@AIToolsforLawyers Instagram: https://www.instagram.com/aitoolsforlawyers [email protected]

SHOW NOTES Every CLE season produces a stack of materials that make sense in the room and mean nothing six months later. NACBA's "Me, Myself, and AI" program is the exception. Four consumer bankruptcy lawyers — all past guests on this show — didn't just talk about using AI, they handed over the actual prompts and skills they run in their own cases. Ron got NACBA's permission to post it all on the website. The question this field note answers isn't whether you should use AI in a bankruptcy practice. It's which of these four entry points is yours. IN THIS EPISODE: - NACBA's "Me, Myself, and AI" program and how the materials ended up free on our website: https://lawyeraitoolkit.com/deliverables/#row-WoaNrpcH4Y - Jenny Doling's 30-prompt consumer bankruptcy library, organized around the actual workflow: intake, means testing, schedules and petition prep, Chapter 13 plan work, trustee issues, 341 prep, reaffirmations, exemptions - Jenny Doling's firm-built AI skills: a pay stub analyzer, a bank statement analyzer, a debt analyzer, and a tax return analyzer - What the bank statement analyzer actually outputs — an Excel worksheet plus a separate attorney work product memo - Tara Salinas's argument for solving one annoying problem instead of building an AI strategy - Tara Salinas's "Your First Closed File" exercise — a 30-minute AI test run on a case you already know the answer to - Chad Van Horn's shift from prompts to skills, and the four questions every skill needs to answer - Chad Van Horn's petition quality control and pleading reviewer examples - Matt McCune's voice, video, software roadmap for building AI into a practice from the ground up - The NACBA50AI discount code for non-member listeners WE ALSO DISCUSS: - Jenny Doling's document renaming, missing document analysis, email organization, and fee quote builder tools - Tara Salinas's point that you don't need a tech committee — if you can explain an assignment to a good paralegal, you can start - Chad Van Horn's rule: if you've corrected AI twice, stop arguing with it and fix the instructions - Ron's own history with NACBA, going back to a 2012 conference in San Antonio DOWNLOAD: NACBA's "Me, Myself, and AI" Materials Four consumer bankruptcy lawyers posted the actual prompts, skills, and workflows they use in their own practices — not slides, tools. Covers: - Jenny Doling's 30-prompt bankruptcy library and her firm's AI skills (pay stub, bank statement, debt, and tax return analyzers) - Tara Salinas's "Your First Closed File" 30-minute exercise, prompts included - Chad Van Horn's four-part framework for turning a prompt into a repeatable firm skill - Matt McCune's voice, video, and software roadmap Also: NACBA is offering podcast listeners $50 off a new membership with code NACBA50AI, through October 31. KEY TAKEAWAY These aren't hypothetical use cases. They're what four working bankruptcy lawyers actually run in live cases, documented well enough that another lawyer can use them tomorrow. That's the difference between a CLE slide deck and a toolkit. Jenny's build gives Jetsons lawyers a full AI-enabled workflow to study. Tara's "just fix one problem" approach is built for Simpsons lawyers who've dabbled but never found a starting point. And Chad's rule — correct it twice, then fix the instructions — is exactly the governance discipline every Flintstones lawyer will need the moment they stop resisting and start experimenting. MENTIONED IN THIS EPISODE: - NACBA (National Association of Consumer Bankruptcy Attorneys) - "Me, Myself, and AI" (NACBA program) - Jenny Doling - Tara Salinas - Chad Van Horn - Matt McCune - Wispr Flow  - HeyGen - NACBA50AI ($50 discount code - Good until October 31, 2026) Highlight Reels on YouTube and Instagram [email protected]

SHOW NOTES Lawyers are building Claude skills, custom GPTs, and homegrown intake bots faster than ever. Most of what they're building will be obsolete in six months, because the foundation models keep swallowing whole product categories. Sabih Siddiqui left Big Law to co-found Irys.ai, a general-purpose legal AI platform. What started as a $299 off-the-shelf app turned into something closer to a custom infrastructure shop. His view is that a skill is a feature and a platform is a different thing. Should you build your own AI tools, buy them off the shelf, or pay someone to build them for you? In this episode: Why Sabih thinks lawyers are "behind" on AI, and why the conference rhetoric hasn't moved past ROI and basic privacy fearsThe sustainability problem with building on Claude skills when the platform underneath changes every weekAn in-house legal team that ran its intake on Claude skills, and what changed when Irys rebuilt it with 36 engineering hoursIrys's "dirty little secret": the $299 off-the-shelf app is an entryway into custom, white-labeled buildsHow a two-attorney pilot grew to 77 users across legal ops, compliance, and financeGrowing to roughly 400 law firms and in-house teams through word of mouth, Facebook groups, and bar association channels, with no marketing spendHarvey and Legora as first movers, and what Sabih calls "wave 2" of legal AIFlat-fee opinion letters at $3,500 that take one hour, and why clients are paying for decades of judgment rather than the hourIn-house clients demanding proof of how outside counsel uses AI, including shared workspaces and per-client-folder ROI reportsRon and Heather's disagreement over whether rank-and-file lawyers want custom builds or something that simply works off the shelfWe also discuss: Every CLE is about AI now, including Ron's skepticism about who's attending the Subchapter V sessionHeather's pushback that lawyers are racing to the AI CLEs, not implementing anything back at the officeTools versus platforms: whether a firm gets better ROI from mastering Word, Slack, and the phone system or from adopting an AI platformPractice Signals: a PI firm on Reddit losing track of who owns the next step on new cases. Sabih's answer is to set it up in Claude or ChatGPT if you don't want a vendor.Claude for Legal versus OpenAI's legal offering, and why Sabih thinks features are commoditizing while workflows hold the valueToken-based billing and what happens when two attorneys run up a Claude bill in two weeksKey Takeaway Sabih's sharpest point is that a feature isn't infrastructure. A Claude skill gives the model a layer to work on, but it does nothing about ingestion, memory, permissions, or the three systems your documents actually live in. Every builder has to decide whether they're building something durable or something the next model release will make irrelevant. The judgment behind the work stays with the lawyer. Everything else is up for grabs. That split runs straight through the Simpsons crowd. The lawyer who's already building skills at night is on a Jetsons path and needs to know where the ground is stable. The lawyer who shows up for a 30-second hearing and goes back to the office doesn't want to build anything. He wants something that works for $299 a month and someone who answers the phone when it doesn't. Heather and Ron don't agree on which group is bigger, and neither answer is obviously right. Mentioned in This Episode Sabih Siddiqui, CEO & Co-Founder, Irys (irys; formerly Akita / Iqidis)HarveyLegoraClaude, Claude skills, Claude Code, Cowork, Claude for Legal (Anthropic)ChatGPT and OpenAI's legal offering Eve LegalLexis (Harvey integration)Palantir (the "Palantir approach")SkaddenGmail, Google Drive, Slack, Microsoft Teams, Zoom, Skype, Microsoft WordTTAB litigationHighlight Reels on YouTube and Instagram [email protected]

SHOW NOTES Every lawyer who's ever complained that no software vendor understands their practice has secretly wondered the same thing Tara Salinas actually did something about: what if you just built it yourself? Tara, a consumer bankruptcy lawyer in Denver, didn't buy a client intake system — she built one, on a stack most lawyers have never heard of, and she's convinced you could too. If you've never touched code, does "build it yourself" actually mean anything for your practice — or is it Jetsons talk that only works for people who were already halfway there? In this episode: - How Tara went from a folder of sideways phone photos labeled "scan001" to a Claude-built client portal that auto-identifies, renames, and sorts every document a client uploads - The nuts and bolts of that portal: a Supabase database on a Vercel/Next.js front end, running an API that scores its own confidence and routes anything uncertain to a "needs a human eye" folder - Why Tara moved from ChatGPT to Claude in February 2026 and started building with Claude Code, Projects, and custom skills - ABA Formal Opinion 512 and the argument over what actually counts as a "self-learning" AI tool requiring client disclosure and consent - Individual vs. team AI subscriptions — why sharing one login with staff likely isn't enough to meet the compliance bar, regardless of platform - The difference between a "skill" and an "agent" — Ron's Family Guy arms analogy, and where agentic activity starts creating real risk - Why AI chatbots are steering financially distressed consumers toward debt settlement instead of bankruptcy, and what that means for the profession - NACBA's incoming AI Advantage program: a prompt library, a skill library, and a microlearning track for training the next generation of bankruptcy lawyers - Tara's client AI policy — the privilege risk of running case documents through a third-party AI tool, and why she bills clients who hand her "AI slop" - The Users and Builders framework: what a user does with a business problem versus what a builder does with the same one We also discuss: - Tara's path from a near-miss patent law career and an unfinished electrical engineering degree into bankruptcy law - The bankruptcy bar's demographic gap after 15 quiet years under BAPCPA, and why the profession needs more young lawyers now - Practice Signals: a burned-out seventh-year associate on Reddit weighing solo practice against leaving law to sell AI software - How a bankruptcy lawyer can go solo with volunteer cases, trade groups, and a notoriously collegial bar - Screenshot hoarding and how to actually manage the AI tool tips piling up on your desktop - AI as "durable infrastructure" — Ron's comparison to the internet and the electrical grid Key Takeaway Tara Salinas didn't set out to become a software developer. She set out to stop losing hours to a folder of unnamed scans, and one afternoon with ChatGPT turned into a purpose-built client portal running its own confidence-scored document API. The lesson isn't that every lawyer needs to learn Supabase and Next.js — it's that the line between "using a tool" and "building one" is a lot thinner than most lawyers assume, and crossing it starts with one file, not a development budget. This lands differently depending on where a lawyer sits. Flintstones lawyers need permission just to try the one file Tara keeps preaching. Simpsons lawyers are the ones actually deciding whether to stay users forever or start building small — a bank statement analyzer, a skill, a prompt library entry. Jetsons lawyers like Tara are already past that question, and the show's real pivot this season is that Users and Builders, not Flintstones and Jetsons, is becoming the more useful way to sort the room. Mentioned in This Episode - Tara Salinas, consumer bankruptcy lawyer, Denver, Colorado - NACBA — Me, Myself, and AI webinar; incoming AI Advantage program - ChatGPT - Claude, Claude Code, Claude Projects and skills - ABA Formal Opinion 512 (2024) - Supabase, Vercel, Next.js - Glade.ai - Yvonne Nath and last week's episode on Counsel Commons and skills vs. agents - Jen Grondahl Lee, Lawyers Success Network - Complete Bankruptcy/Team Accelerator (Ron's referenced training program) - Family Guy (referenced for the skills-vs-agents analogy) - ECF and Infusionsoft (referenced in Ron's email-automation story) - DocuSign [email protected] Highlight Reels on YouTube and Instagram

SHOW NOTES A legal industry business professional spends her career building expertise, then watches AI automate the most valuable parts of that work. That's not a hypothetical — it's what this week's guest proved could happen. Yvonne Nath, founder of Counsel Commons, joins Ron and Heather to talk about a marketplace where lawyers and legal professionals can buy and sell the AI prompt libraries, skills, and agents they've built. But the conversation goes somewhere bigger: what happens when AI doesn't just make legal work faster, but cheap enough to create an entirely new market for the millions of people who can't afford a lawyer today?    In this episode:    - Yvonne Nath's path from law firm operations and strategy work to founding Counsel Commons, a marketplace where lawyers and legal professionals sell the AI automation tools they've created - How Yvonne productized most of her own   law firm management expertise in about a month, and what that says about the future of legal ops roles - Some distinction between skills and agents, and why Yvonne deliberately builds almost none of the latter - How Counsel Commons splits into two channels — bar-verified attorneys and paralegals versus business professionals — to manage unauthorized-practice-of-law risk - Counsel Commons’ vetting process: AI review for prompt injections and input/output accuracy rather than manually testing every submission - The pricing model for sellers — a flat listing fee for legal professionals who keep nearly all the revenue, versus a revenue split for business-side tools - Yvonne's decision to run her own automations on local, open-weight models rather than through Anthropic or another cloud provider, and the data-privacy reasoning behind it - The $1,500 divorce and the access-to-justice math — why the millions of people priced out of traditional legal help represent an untapped market, not a threat to existing practice - Ron and Yvonne's disagreement over whether productizing legal services cannibalizes a firm's existing client base or reaches an entirely different one - What Yvonne means by an "AI native" law firm, and why she thinks real transformation requires starting from scratch We also discuss: - Counsel Commons' review system, built around objective performance questions rather than star ratings - Heather's team meeting encouraging her paralegals and virtual assistants to build AI literacy now, not later - Ron's case that solo lawyers would get further mastering Word, Excel, and their phone systems than chasing AI tools haphazardly - This week's Practice Signal: a solo domestic-law attorney wondering whether they now need to hire a consultant just to use tools that were supposed to make everything easier - The NACBA webinar where consumer bankruptcy attorneys discussed building their own Claude skills Key Takeaway Yvonne's real point isn't that AI makes legal work faster — it's that legal work becomes a different business (and, hopefully, a more rewarding one) once it's cheap enough for ordinary people to actually buy. A $1,500 divorce or a $50 landlord dispute isn't a discount version of a $20,000 case. It's a client who was never walking through the door in the first place. That lands hardest for lawyers still treating AI as a way to shave hours off existing work rather than a way to reach people who currently have no lawyer at all. It's also this season's Users and Builders split in miniature: Yvonne builds, verifies, and sells; most solo and small firm lawyers just want something proven that works off the shelf — and she thinks that gap is exactly the market Counsel Commons should serve. Mentioned in This Episode - Yvonne Nath, founder of Legal InnovAI - Claude - Qwen the local, open-weight model Yvonne runs on her own hardware - Lawvable / Lawv.ai — an open-source skills library Yvonne referenced as possibly connected to Claude's built-in legal skills - Stripe - NACBA (National Association of Consumer Bankruptcy Attorneys) - divorce.com - Upsolve [email protected] Highlight Reels on YouTube and Instagram

SHOW NOTES A lawyer running a massive document production doesn't read every page anymore — she can't, and nobody expects her to. So when does that stop being lawyering and start being blind faith? Jim Casey, Chief Privacy Officer at Akamai Technologies, joins Ron and Heather to argue there's a real answer, one that depends entirely on whether a bad AI output can still be caught before someone acts on it. When is an AI mistake actually reversible, and when have you already handed your judgment to the machine? In this episode: Jim Casey's framework for deciding when a lawyer can trust an AI-generated answer, drawn from his two published pieces On Knowledge in An Agentic World and Agentic Workflows in ComplianceThe "Wilma Simpson" hypothetical — a bankruptcy lawyer using AI to summarize a 180-page court order, and where the real risk actually sitsRon and Jim's sharp disagreement over whether a "reversible" AI mistake stays reversible once a client acts on the lawyer's advice built from itWhy reviewing a massive document production (sorting, summarizing, deduplicating) shifts trust from individual verification to system-level designAn e-commerce company's real-world case study training a model on product-safety complaint data — six months of retraining to reach 90–95% accuracyThe malpractice waiver hypothetical: what an in-house lawyer does when outside counsel wants to disclaim liability for AI-assisted workClient consent and data privacy limits on using AI to process a client's own personal informationThe difference between "human in the loop," "human on the loop," and "human out of the loop" — a framework borrowed from autonomous weapons policy"Compliance theater" — why an audit log nobody actually reviews isn't a safeguardThe "non-delegable duty to think" — what AI can never take off a lawyer's plateWe also discuss: The Flintstones/Simpsons/Jetsons framework applied to matching lawyers with the right clientsA Reddit-sourced Practice Signals case: a kind, hardworking associate who struggles with vocabulary and comprehension — does AI help, or just hide the problem?Chris Ryan's AI training tool for new associates, Benchsim.ai, now SOC 2 compliant and video-capableThe Season 3 shift from Flintstones/Simpsons/Jetsons to a new "Users and Builders" frameworkFirst tips for builders and users: "make failure visible" and "ask and ask again"Key Takeaway Jim Casey's real point isn't that lawyers can or can't trust AI — it's that trust isn't one setting you flip on. The question changes at every step in the chain: is this output feeding the lawyer's own thinking, or is it about to become the advice a client acts on? Reversibility disappears the moment a lawyer hands that answer downstream. Everything else — audit logs, compliance sign-off, vendor demos — is theater unless someone is actually positioned to catch the mistake before it moves. This matters most for Simpsons lawyers who've adopted a tool but haven't thought through where accountability actually sits in their own workflow. Flintstones lawyers aren't there yet, and Jetsons lawyers like Jim are already building the guardrails. But the solo or small firm lawyer running document review or drafting client advice off an AI summary needs a concrete answer to Jim's closing question: what do you do with the confidently wrong answer? Mentioned in This Episode Jim Casey, VP & Chief Privacy Officer, Akamai TechnologiesJim Casey's articles: Knowledge in an Agentic World and Agentic Workflows: A New Frontier for ComplianceChatGPTClaudeHarvey (AI legal tool)Legora (referenced as an enterprise legal AI tool)Chris Ryan and his AI training tool for new lawyers (previously featured on the show)Reddit (source of this episode's Practice Signals story)Highlight Reels on YouTube and Instagram [email protected]

SHOW NOTES Chad Van Horn built an AI petition review process that cut his firm's amendments by 80 percent. He also thinks AI is making lawyers dumber. Both are true in the same law firm, and figuring out how to hold both truths at once is the whole game. If AI can teach a new associate everything a senior partner knows, what happens to the judgment that only comes from doing the reps yourself? In this episode: - How Chad's Claude-built bank statement analyzer replaced his highlighter-and-manual-review process - The "volunteers, not hostages" approach to rolling AI out across a hundred-person firm - Why Chad's AI-powered petition review step reduced amendments by 80 percent - How Chad uses himself as a human "AI agent" to bring reluctant staff along without forcing adoption - What Chad's firm built with GLADE, its all-in-one bankruptcy practice management platform, and how Claude and GLADE connect directly - Why Chad believes AI is making lawyers less capable critical thinkers, even as it makes them more efficient - How Chad trains AI skills to check judges' local rules, procedures, and order formatting, and updates them when orders get kicked back - Why bankruptcy filings are rising sharply and what that means for firm capacity - How Chad and Heather debate what percentage of bankruptcy work is really legal judgment versus process - Chad's prediction that practitioners who don't adopt AI may eventually be unable to compete We also discuss: - The gap between Gemini's early data advantage and its current market position - Whether token pricing stays this cheap once AI vendors need to turn a profit - Vendor reliability risk as new AI legal tools flood the market - The FSJ framework applied to Season 2's document-by-document AI onboarding approach - This week's Practice Signal: a solo practitioner asking an inexperienced legal assistant to build the firm's entire tech stack - Florida's vehicle exemption and how a small asset detail can cost a client their car Key Takeaway Chad's petition review process shows what disciplined workflow gets you: fewer amendments, tighter output, and a documented feedback loop that trains the humans as much as it trains the AI. But Chad's own warning is louder than his results. The same tools that make firms faster are, in his view, quietly eroding the critical thinking lawyers need to catch the case that doesn't fit the pattern. For Flintstones lawyers, GLADE and Chad's petition reviewer are proof that meaningful adoption doesn't require a hundred-person firm to pull off. For Simpsons lawyers building their own skills, Chad's volunteers-not-hostages approach and his habit of weekly skill refinement are a directly usable model. For Jetsons lawyers already running full systems, Chad's warning about critical thinking is the harder, more uncomfortable takeaway. Mentioned in This Episode - Claude - ChatGPT - Gemini and Google Workspace - GLADE, practice management and bankruptcy software (name as heard on the recording, confirm spelling) - Clio - Litify (referenced on the recording as "Lytify," confirm intended name) - Salesforce - HubSpot - Easy Filing - QuickBooks - Best Case - CoCounsel (Westlaw) - David AI - NACBA's "Me, Myself, and AI" webinar - Jenny Doling - Van Horn Law Group - Chad Van Horn, author of The Debt Life and Everything You Need To Know About Bankruptcy in Florida Highlight Reels on YouTube and Instagram [email protected]

SHOW NOTES A federal bankruptcy judge just told us she doesn't personally use generative AI in her own chambers - and that's exactly why this conversation is worth your time. Judge Elizabeth L. Gunn, who sits on the U.S. Bankruptcy Court for the District of Columbia, spent nearly an hour with Ron and Heather working through what AI actually changes about the practice of law, and what it doesn't. Central question: If AI can now do in seconds what used to take a junior associate hundreds of hours, who trains the next generation of lawyers - and who's accountable when the machine gets it wrong? In this episode: - Why Judge Gunn deliberately limits AI use in her own chambers, to protect it as a training ground for law clerks and interns - How judicial opinions on AI vary widely across the bench - "as many opinions as there are black robes" - Whether it matters, evidentiarily, if a summary took an associate 700 hours or an AI tool 700 seconds to produce - How firms could develop internal AI tools for chapter 11 practice, and what foundation they'd need to lay for the court - Why Judge Gunn believes generative AI is hollowing out the mid-level associate pipeline - How fee applications are starting to raise questions about AI subscriptions and engagement agreements - Judge Gunn's real experience with hallucinated cases in her courtroom, including sanctions and a running database of violations - Why she believes a sliding scale of culpability makes sense for different kinds of hallucinations in legal briefs We also discuss: - Whether courts should look behind a flat-fee agreement to question consumer bankruptcy AI tool charges - How Lexis and Westlaw brief-checker tools are used in chambers to catch bad citations - Why Judge Gunn is skeptical the UAE's centralized, nationwide AI judiciary approach would work in the U.S. federal system - Differing philosophies among law schools on AI in the classroom, from banning electronics to teaching practical AI skills - Whether it's the court's role to inquire into AI-related terms of an engagement agreement when reviewing a fee application Key Takeaway: Judge Gunn's core message is that AI doesn't require a parallel set of rules. Rule 9011, the rules of evidence, and existing duties of competence and candor already cover it. What changes isn't the standard - it's how carefully lawyers have to apply it, and how quickly courts have to learn to tell the difference between an inaccurate pin cite and an argument built on a case that doesn't exist. For Flintstones lawyers, that's reassurance: you don't need to master new technology to stay compliant, because your existing ethical obligations already cover AI-created work product. For Simpsons lawyers dabbling with tools like Claude or ChatGPT, Judge Gunn's evidentiary framework is a warning - if you can't establish the foundation for how a summary was created, speed doesn't save you. For Jetsons lawyers building AI into chapter 11 workflows, her comments point to exactly what a judge will want documented before she trusts the output. Mentioned in This Episode: - Judge Elizabeth L. Gunn, U.S. Bankruptcy Judge, District of Columbia - Bad Boys of Bankruptcy podcast (Judge Gunn's podcast) - Claude - ChatGPT - Harvey - Lexis - Westlaw - Best Case (bankruptcy case management platform) - Rule 9011 - Northern Pipeline Construction Co. v. Marathon Pipe Line Co. - Stern v. Marshall - Nancy Rappaport, UNLV School of Law - University of Chicago School of Law - New York and Florida AI ethics rules - DC Bar ethics rule on AI HIGHLIGHT REELS YouTube: https://www.youtube.com/@AIToolsforLawyers Instagram: https://www.instagram.com/aitoolsforlawyers [email protected]