Three Takes on AI

This episode of Three Takes on AI uses the Musk v. Altman trial (with Microsoft as co-defendant) as a springboard to challenge the binary "Star Trek vs. Terminator" framing of AI and ask the harder question: who is actually accountable when AI fails.

The hosts examine three competing AI worldviews exposed by the courtroom drama: Musk's safety-first existential-risk stance (undercut by xAI training Grok on OpenAI's models), Altman's view of AI as an unprecedented positive force requiring commercial scale, and Microsoft's platform-shift play that uses Copilot to own the distribution layer while others fight over models.

• Brian Silverman takes the point that AI doesn't "go rogue" — it behaves exactly as it was trained, and the "rogue" label conveniently shifts blame away from the humans who built and deployed it.

• Campbell Robertson challenges the industry's language problem, arguing that calling production errors "hallucinations" softens a hard problem that carries real consequences and demands real ownership.

• Michael Muhlfelder points out that the governance controls missing from today's AI deployments are precisely what experienced business and IT professionals spent decades building.

The episode pushes past Musk's own utopia-or-apocalypse framing to focus on responsibility, accountability, and governance — continuing the show's ongoing thread from Shadow AI and the MIT ROI study to agentic risk and now this landmark trial.

The Trial as BackdropThe Three TakesThe Through-Line

AI governance is like a self‑driving car that must keep humans firmly in the loop. In this episode, the hosts argue that when we “forget the human,” AI failures stop being about bad mortgages and start threatening real lives and health.

They explore why every powerful AI system needs a clear challenge mechanism that quickly routes questionable decisions to a human who can correct or confirm them, rather than letting automated errors silently pile up. Using the metaphor of a car that not only drives itself but constantly monitors the driver, they show how governance should continuously check human engagement instead of assuming oversight is happening. The conversation closes by stressing that responsible AI is ultimately about preserving human agency and safety, and they tee up the next installment of their “AI Governance: Ignorance and Satire” series in two weeks.

On the same day NASA's Artemis II crew flies around the moon — a triumph of human courage and intellect — we tackle a far less celebratory reality: what happens when the AI you championed at work comes for your job?

From Oracle's massive 30,000-person layoff to Atlassian replacing its CTO amid AI restructuring, no role is safe — entry level to the C-suite. We unpack the AI Champion Paradox (deploying AI only to be displaced by it), the rise of AI washing (using AI as cover for cost-cutting), and the death of the knowledge worker as Peter Drucker defined it 76 years ago.

We also dig into agentic AI, digital twins, the security risks of building your own AI replica, and the economic ripple effects when one company's decision wipes out $4.5 billion in annual wages overnight.

But it's not all doom. Inspired by four astronauts aboard Artemis II proving that humans still choose to do hard things, we explore what's uniquely human, how to shift from knowledge work to judgment work, and practical steps — whether you're a recent grad, a mid-career professional, or a seasoned exec — to adapt and stay relevant.

🎙️ Hosts: Brian Silverman, Michael Muhlfelder & Campbell Robertson
⏱️ Runtime: ~22 min

Three Takes on AI — Where business, technology, and implementation meet AI. New episodes every other Tuesday.

#AI #FutureOfWork #AIGovernance #KnowledgeWorker #AgenticAI #Artemis2 #ThreeTakesOnAI

Everyone's rushing to use AI for everything — but when should you actually hold back? In Episode 11 of Three Takes on AI, the crew flips the script and tackles the scenarios where AI isn't the answer. From missing governance frameworks to over-reliance on AI outputs, we break down the real risks of using AI without the right guardrails. Three perspectives, one essential question. Subscribe at threetakesonai.com and share your take.

The episode explores the question of accountability when AI systems make consequential decisions. The hosts discuss who bears responsibility — individuals, organizations, or AI companies — when AI takes actions that go wrong. The conversation was sparked by a letter written by a friend named Sheila, intended for the CEO of Anthropic, addressing data sovereignty and AI sovereignty — specifically, how individuals can understand where and how AI-driven decisions are made.

Would you trust AI to diagnose your child's illness? From a revealing encounter with the CIO of Federal Prisons to the real-world gap between intelligence and experience, Brian, Campbell and Mike dig into why human judgment still holds an edge that algorithms can't replicate — and what we risk losing if we forget that.

AI and Human Intelligence are different but how should they work together? Is human intelligence at risk from AI? How do we govern AI and HI together for an organization. We discuss how a fine balance needs to be incorporated into business strategy and planning to augment humans with AI and not the other way around.

After all we humans built AI and we need it to serve our needs and wants for a better planet and a productive working environment.

Shadow AI: The Workforce Revolution Hiding in Plain Sight

Your employees are already using AI—whether you know it or not. Studies show 59% of workers hide their AI use from their bosses, and 93% have shared confidential company data through unauthorized tools. Is this a security nightmare or the greatest source of untapped innovation in your organization?

In this episode, we explore Shadow AI—the unsanctioned use of AI tools like ChatGPT that’s become the modern equivalent of the workplace “cheat sheet.” We break down:

• Why Shadow AI often delivers better ROI than formal enterprise AI initiatives

• The real security risks (Gartner predicts 40% of enterprises will face Shadow AI breaches by 2030)

• Lessons from Samsung’s infamous data leak—and why that leaked code can never be retrieved

• Whether Shadow AI is an IT problem, a cyber problem, or a business problem (spoiler: it’s all three)

• Practical strategies to turn grassroots AI experimentation into company-wide best practices

Shadow AI isn’t going away—46% of employees say they’d keep using AI even if banned. The question isn’t how to stop it. It’s how to learn from it.

In 2026, agentic AI finally has to grow up. With analysts projecting that roughly 40% of enterprise applications will embed task-specific AI agents by this year, the gap between glossy demos and dependable production systems is about to get brutally obvious. Join hosts Brian, Mike, and Campbell as they unpack why 2026 is the make-or-not year for AI agents, what will separate durable platforms from "agent-washed" automation, and how smart organizations can turn governance and accountability into a competitive advantage rather than a brake on innovation.

Brian, Campbell, and Mike wrap up the year with their Three Takes on AI for 2025, comparing what actually happened to the trends they predicted, calling out the biggest surprises, and sharing where they think AI reality is headed next for leaders, builders, and operators.