197: Jason Robinovitz: "What Happens When Students Outsource Their Thinking To LLMs?"

199: "How Do We Actually Get ROI From AI?" (reflections on Lauren Cappell)

🧠 Erik’s Take After speaking with Lauren Cappell, Erik walked away with a realization that extends far beyond the legal profession: the biggest barriers to successful AI adoption aren’t technical—they’re human. While most organizations are focused on finding the right tool, Lauren challenged a deeper question: Do leaders actually understand the work they’re trying to automate? The conversation highlighted how AI is exposing weaknesses that have existed inside organizations for years—unclear processes, underdeveloped systems, and leadership teams that haven’t fully mapped how value is created. AI doesn’t eliminate those problems. It magnifies them. 🎯 Top Insights from the Interview 1. You Can't Automate What You Don't Understand Before AI can improve a process, leaders need clarity on what actually happens inside that process. Most work isn't a straight line. It's a web of decisions, judgment calls, handoffs, and exceptions. Mapping that reality is difficult—but it's the prerequisite for meaningful automation. 2. AI Adoption Is a Leadership Challenge The organizations that succeed with AI won't necessarily have the best technology. They'll have leaders willing to: *  Invest time understanding workflows  *  Train their people to use AI effectively  *  Build systems for oversight and quality control  *  Commit resources to ongoing learning  Technology isn't the limiting factor. Leadership engagement often is. 3. Efficiency Can Create a Talent Problem Many professions develop expertise through repetitive, foundational work. When AI removes that work, organizations gain efficiency—but may lose an important training ground for future experts. The challenge becomes identifying: *  What skills actually need to be developed  *  Which activities are valuable learning experiences  *  How simulations or alternative training methods can accelerate mastery  4. Human Oversight Isn't Going Away One of the biggest misconceptions about AI is that it removes people from the process. In reality, many workflows require human intervention at critical checkpoints. Leaders still need people who understand the work well enough to recognize mistakes, redirect efforts, and improve systems over time. 5. The Next Step Matters More Than the Perfect Plan Companies waiting for the perfect AI solution may be making the biggest mistake of all. The organizations gaining advantage today are experimenting, learning, adapting, and building capability one step at a time. Five years from now, experience will compound. 🧩 The Personal Layer One idea particularly resonated with Erik: the realization that he had been viewing AI adoption primarily as a process problem. His focus had been on understanding workflows well enough to automate them. Lauren pushed that thinking further. Before organizations can automate effectively, leaders must invest in creating people who are capable of working alongside AI. That means training, literacy, experimentation, and ongoing development. The real bottleneck may not be process mapping. It may be whether leaders are willing to make the investment required to build an AI-capable workforce. 🧰 From Insight to Action If you're exploring AI inside your organization, consider these questions: *  Can your team clearly explain how a critical process works today?  *  Where does judgment—not just execution—create value?  *  Which repetitive tasks are teaching future experts important skills?  *  How much time are you investing in AI literacy and training?  *  What is the next practical step you could take instead of waiting for a perfect solution?  The goal isn't to automate everything. The goal is to become the kind of organization that learns how to leverage AI better every year. 🗣️ Notable Quotes "You can't automate what you don't already know how to do." "The real hurdle isn't technology. It's leadership." "If AI does all the grunt work, how do we develop the next generation of experts?" "The companies figuring out AI today will be far ahead of the companies waiting for AI to figure itself out." "The next step is the most important one." 🔗 Links & Resources * Listen to Lauren Cappell's Episode [https://podcast.languageofleadership.io/201-lauren-cappell-time-kills-deals-how-faster-legal-work-creates-real-revenue]

198: Lauren Cappell: "Time Kills Deals: How Faster Legal Work Creates Real Revenue"

Lauren Cappell and Erik unpack why realizing ROI from AI adoption in law is harder than it sounds, especially under legacy business models. Lauren argues the real shift is not just efficiency, but value creation: new kinds of work, faster cycles, better workflows, and better pricing that aligns incentives. They also dig into the training and QA demands required to ensure AI outputs are trusted. 👤 About the Guest Lauren Cappell is a strategist at the intersection of enterprise, law, and artificial intelligence. With leadership experience at Amazon, Thomson Reuters, and BlackBerry, plus startup roles, she helps organizations translate AI capability into measurable business value. Her mission is to eliminate busy work and replace it with systems that elevate human potential. 🧭 Conversation Highlights * Billable hours create tension because AI can reduce the time needed to complete a task, but it also enables work that was previously uneconomic or impossible. * In law, ROI shows up differently across in-house teams, large firms, and smaller firms, because incentives and constraints vary. * AI services that offer usage based billing to legal teams are the right move right now: they can assist in driving adoption and ensure aligned incentives between the AI service and the buyer/user. * Adoption depends on training, human in the loop QA, learning how to work with AI outputs, and rethinking the way we work, not just training on how to use individual AI tools. 💡 Key Takeaways * AI ROI has to be tied to reduced delivery cost and/or increased value, and it often requires work redesign rather than speed alone. * Billable hours are not the whole story, because much of legal work happens outside large firms and without the same pricing constraints. * Usage based billing can align incentives , but it increases the importance of effective training and successful “usage” definitions. * The primary limiting factor to seeing value from AI investments is people and process: automation requires you to know the workflow well enough to de-risk it and to QA output quality. ❓ Questions That Mattered * How does a billable hours model persist when AI reduces the time needed for certain tasks? * Where does ROI show up first across in-house teams, large firms, and smaller firms? * Why is adoption often slower than expected, even for sophisticated AI users? * What does it mean in practice to deploy agents or automation while staying accountable for quality and decision making? 🗣️ Notable Quotes * “Time kills all deals.” * “If you’re waiting for your clients to ask you about AI usage or to challenge you on it, I don’t think that’s a great position to be in.” 🔗 Links & Resources * Follow Lauren Cappell's LinkedIn [https://www.linkedin.com/in/lauren-cappell/] * Check out Lauren’s personal website: deathofbusywork.com [https://deathofbusywork.com/]

197: Jason Robinovitz: "What Happens When Students Outsource Their Thinking To LLMs?"

Jason Robinovitz explores what education needs to become in an AI-saturated world. He argues that AI will increase opportunities, but only if schools keep teaching students to think for themselves. He shares practical classroom tactics to reduce cheating, why incentives in education often fail learning, and which human skills, like critical thinking and soft skills, will matter most as hiring shifts. 👤 About the Guest Jason Robinovitz is CEO, COO, and General Counsel at Score at the Top Learning Centers, Score Academy, and JRA Educational Consulting. The family-owned organization was founded in 1980 and operates across South Florida. A former medical malpractice attorney, he joined the business in 2008 and brings a legal, systems-minded approach to education.  🧭 Conversation Highlights * AI will not remove the need for education, but it will force schools to verify thinking through process integrity and human judgment. * AI detectors are unreliable, so educators need better proof methods like Google Docs version history and proctored, handwritten exams. * Education incentives can drift toward standardized-test outcomes rather than learning, which drives both cheating and hollow credentials. * Jason emphasizes the essential skills employers want: soft skills and critical thinking, supported by reading, mental math, and discomfort-building practice. 💡 Key Takeaways * The most valuable student skill in an AI world is thinking for yourself, not outsourcing decisions to LLMs. * Cheating controls should prioritize process visibility and classroom verification, not detector trust. * Employers will increasingly hire for soft skills and critical thinking demonstrated under real conditions, not just grades. * Tech in education should augment teachers, especially for feedback, and focus on building “understanding” rather than automating judgment. ❓ Questions That Mattered * What does education need to teach when students can outsource work to AI in minutes? * How can educators ensure evidence of thinking without relying on weak AI-detection tools? * Will prestige-based hiring proxies break down, and what would replace them? * How do we help critical thinking grow under stress, not just in ideal conditions? 🗣️ Notable Quotes * “In my opinion, the most important thing that students can learn today is the ability to think for themselves.” * “Most AI detectors suck. They are full of false positives. They can’t be relied upon.” * “Education is going to become incredibly important, but probably not in the ways that most people are thinking about it.” 🔗 Links & Resources * Follow Jason Robinovitz's LinkedIn [https://www.linkedin.com/in/jasonrobinovitz/]  * Follow Jason Robinovitz’s Substack: News From The Top [https://jasonrobinovitz.substack.com/] * Check out Score Academy’s website: Score-Academy.com [http://score-academy.com]  * Check out JRA Educational Consulting: jraEducationalConsulting.com [http://jraeducationalconsulting.com]  * Check out Score At the Top’s Website: ScoreAtTheTop.com [http://scoreatthetop.com]

196: "Agent Collaboration Should Look Like Co-Working, Not Hand-Offs" ft. Justin Coats

Erik and Justin talk through why AI agents get stuck when teams cannot articulate how work is done, and why the answer is iterative agent deployment with guardrails, sandbox testing, and ongoing process refinement. 🧭 Conversation Highlights * Most professionals struggle to explain the steps of their work, which blocks automation and agent adoption. * Agents feel hard to hand off to because people fear losing control, making mistakes, or looking foolish. * A practical path forward is iterative process creation: give an agent the goal, tools, and guardrails, then tighten the workflow based on outcomes. * To make iteration safe, teams need preview or sandbox testing plus limits on real-world actions and token budgets to avoid runaway usage. 💡 Key Takeaways * AI adoption is shifting from “AI literacy” to “how to build and govern agents,” so companies need shared understanding beyond IT. * Iterative deployment works better than trying to hard-code every step upfront, but it requires verification checkpoints and process feedback loops. * Sandbox or preview environments are critical for low-risk learning before enabling agents to take real actions. * Token spend should be treated as governance: set budgets and limits per user/team, and track usage in a way leaders can understand. ❓ Questions That Mattered * How do we automate work we cannot clearly describe step-by-step without stalling adoption? * What guardrails let humans feel safe handing tasks to agents, including security, safeguards, and action approvals? * Is there a practical way to test agent workflows in “failure-free” conditions before going live? * When iteration causes back-and-forth, how should companies think about token efficiency and budget controls? 🗣️ Notable Quotes * “It’s hard to automate what you don’t already know how to do.” * “The majority of people have a really hard time articulating precisely how they perform a task.” * “Understanding the tool and technology, and knowing what it can and cannot do, really creates a safer environment.” * “It’s taking our typical structure. You watch how they work and adapt to that.” 🔗 Links & Resources * Listen To Other Episodes Co-Hosted With Justin [https://www.google.com/url?q=https://podcast.languageofleadership.io/categories/i-have-some-ai-questions-with-justin-coats/&sa=D&source=editors&ust=1783541882324986&usg=AOvVaw3cPDGHMfUE2VtEJraxNcHR]

195: Rocky Batzel: "What It Means To Be Ready For Manufacturing At One Million Units Per Month"

Rocky Batzel, inventor and CEO of Snapslide, shares how a decade of tinkering became a child-resistant pill bottle closure designed for one hand and for people with arthritis or other limitations. From the original “aha” at a liquor store to prototyping, patents, and certification testing, Rocky explains the path to commercial viability and what scaling manufacturing for over a million units per month means next. 👤 About the Guest Rocky Batzel is the inventor and CEO of Snapslide. He left medical school and built a career around product invention, focusing on tangible solutions. Snapslide creates a new approach to child-resistant openings for medication containers, aiming for accessibility without sacrificing safety requirements. 🧭 Conversation Highlights * Snapslide’s core idea: replacing torque-dependent bottle caps with a linear, two-stage opening that can be done with limited dexterity * How Rocky shifted from identifying everyday “pain in the butt” problems to searching for prior art, patents, and manufacturability * The business and regulatory gauntlet: child-resistant testing, USP permeation testing, iterative tooling, and certification timelines * How the team is preparing to scale manufacturing and capacity for large pharmacy distribution while continuing to develop OTC variants 💡 Key Takeaways * Accessibility is not a “nice to have.” It is a design constraint that must be baked into safety products from the start. * Great invention is less about finding a new problem and more about observing a familiar problem from a different angle. * Commercial success requires solving for manufacturability, cost, certification, and distribution incentives, not just the mechanism. * Scaling is a timing problem: tooling lead times, capital planning, and facility growth capacity have to align with demand. ❓ Questions That Mattered * What is Snapslide, and what design change makes it usable for one hand or limited dexterity while staying child-resistant? * How do you validate an idea when it is hard to know if you are truly first, or if prior art exists? * What does the child-resistant certification process actually require, including pass thresholds and sample counts? * What keeps you from taking profitable but misaligned deals, and how do you decide what is “worth it”? 🗣️ Notable Quotes * “Simple, which is one of the big barriers to the market.” * “You know it in your gut.” 🔗 Links & Resources * Follow Rocky Batzel's LinkedIn [https://www.google.com/url?q=https://www.linkedin.com/in/rocky-batzel-780a309a&sa=D&source=editors&ust=1780007408071641&usg=AOvVaw36l6gVoGjlRatUhA8iKUI8] * Check out Rocky Batzel’s Company, SnapSlide: www.snapslide.com [https://www.google.com/url?q=http://www.snapslide.com&sa=D&source=editors&ust=1780007408072118&usg=AOvVaw3_vaWKS4HiXCBneCvgP3-R]