Human-in-the-Loop Isn’t Optional: The Real Future of AI in Law with Sam Dixon

What happens when a traditional profession meets generative AI, and now agentic AI? In this episode, I’m joined by Sam Dixon (Chief Innovation Officer) to unpack how law firms can adopt GenAI responsibly, train teams to critically evaluate outputs, and use AI to deliver more value, without burning out in the process.We cover what GenAI is (and isn’t), why humans must stay in the loop, how to overcome resistance to change, the reality of “hallucinations,” and what the next phase, agentic AI, could mean for legal research, negotiation simulation, and early-talent development.Timestamps 0:00 – Intro 1:39 – Sam’s role as Chief Innovation Officer & how he defines innovation5:29 – Getting partners to embrace change (and what actually works)8:39 – Why GenAI adoption is different from past legal tech waves10:26 – Explaining AI to senior leaders (without the jargon)14:49 – “Human in the loop” + the risks of over-trusting AI19:26 – GenAI vs Agentic AI: what changes, and why it matters22:46 – New lawyers, verification skills, and “AI makes people lazy?”33:11 – Simulation: negotiation practice, interviews & training with GenAI47:02 – AI & well-being: tool or threat?What we coverThe real job of innovation inside a law firmHow to get buy-in from busy, skeptical stakeholdersThe difference between AI, ML, GenAI, and Agentic AIWhy hallucinations are misunderstood, and why that mattersGuardrails: verification, accountability, and client expectationsThe future of training: simulations for negotiation, interviews, and client conversationsAI and burnout: productivity vs cognitive load and recoveryInnovation isn’t just tech, sometimes it’s removing a step entirely.GenAI adoption is faster because users can get value immediately, unlike tools that require heavy setup.Agentic AI adds value through multi-step planning, not just answering prompts.“Hallucinations” aren’t a rare bug, GenAI always generates probabilistically; the key is how you manage risk.The winning model is human + AI, where each catches different kinds of errors.As outputs become “more accurate,” verification discipline becomes even more critical, not less.AI may improve efficiency, but well-being depends on how work is redesigned, not just sped up.Guest ResourcesSam Dixon’s Legal Firm: https://www.womblebonddickinson.com/uk/people/sam-dixonSam Dixon’s LinkedIn: https://uk.linkedin.com/in/innovationinlawConnect with Me🌍 Website: https://charlenegisele.com📸 Instagram: https://instagram.com/charlenegisele💼 LinkedIn: https://www.linkedin.com/in/charlène-gisèle-bourliout/✉️ Subscribe to my Life Less Burnt Out NewsletterTimestamps What we coverKey takeawaysGuest Resources