Podcast Summary

Episode Overview

Podcast: Using AI at Work
Episode: 94 — "Using AI vs Human Intelligence: When Should Leaders Trust Machines"
Host: Chris Daigle
Guest: Professor Vasant Dhar (NYU Stern & Center for Data Science, host of Brave New World podcast, author of Thinking with Machines)
Air Date: March 9, 2026

This episode focuses on a critical question for business leaders in the AI era: When and how should executives trust machines over human intelligence? Professor Vasant Dhar draws on his 40+ years in AI to share a pragmatic, richly framed approach to workplace AI adoption, emphasizing risk, trust, and the emerging imperative for “thinking with machines.” The conversation covers AI implementation, failures and success factors in business, trust boundaries, talent concerns, and the coming bifurcation in the workforce.

Main Themes and Purpose

• Trust in AI: Mapping when, where, and why to trust (or not trust) AI in the workplace.
• Business Decision-Making: How leaders can identify AI opportunities, mitigate risk, and avoid common pitfalls.
• Human + Machine Synergy: Why “thinking with machines” is the real future—and why opting out is not an option.
• Strategic Adoption: Insights on “look before you leap” vs. “he who hesitates is lost.”
• Bifurcation of Workforce: The emerging split between those amplified by AI and those left behind.

Key Discussion Points & Insights

1. The Map of Trust: When Should We Trust Machines?

[03:51–09:07]

• Dhar explains his ‘trust framework’ pioneered eight years prior:

  • Trust in AI depends on:
    1. Predictability of the AI’s output: How often is it wrong?
    2. Consequences of error: What happens if it’s wrong?
  • "If a machine is never wrong, you should absolutely trust it. On the spectrum from zero (random) to one (perfect predictability), you map problems based on accuracy vs. cost of error." — Vasant Dhar [04:23]
  • Examples:
    • In finance (low predictability, low consequence): Trust algorithms, even if wrong almost 50% of the time.
    • In healthcare or driverless cars (high risk, severe consequences): Reluctant to trust, even if rarely wrong.
  • This framework produces a “trust heatmap” and an "automation frontier," demarcating when to automate and when human oversight is essential.

• Relates to current generative AI uses:

  • For simple tasks (e.g., checking a basic legal contract): May trust AI.
  • Complex, high-stakes decisions: AI must be verified by humans.
  • "If you’re going to double-check anyway, what good did the AI do you in the first place?" — Vasant Dhar [08:45]

2. Barriers to AI Adoption in Business

[09:07–16:59]

Chris Daigle identifies top executive concerns:

1. Unknown risk—Confusion leads to inaction.
2. Unclear starting point—Where do we meaningfully deploy AI?
3. Lack of trustworthy talent—Who actually knows how to deliver?

• Dhar’s response:
  1. Talent: Don’t hire “wannabes.” Proven successful project experience is essential.
     "You can't just hire talent off the street...you should be skeptical." — Vasant Dhar [10:28]
  2. Where to Start: Failure is often due to poor problem selection, not AI shortcomings.
     • Leadership must own problem selection.
     • Example: Insurance company failed to ask if broker lead-generators were sending higher-risk cases.
     • "MIT study showed 95% of projects failed, not because of the AI, but the wrong problem was chosen...That’s a leadership issue." — Vasant Dhar [11:38]
  3. Assess risk using the trust map: Begin with projects where AI’s accuracy is high and the cost of error is low.

3. Trends and Executive Sentiment: Fear Still Dominates

[18:50–24:03]

• Business leaders are taking AI seriously, seeing it as transformative.

  • "What I'm seeing is serious engagement because this is a pivotal technology. It's here to stay." — Vasant Dhar [18:50]

• Sentiment:

  • “Mostly fear,” both of missing the AI boat and of disruption/failure.
  • Even Wall Street giants (like Morgan Stanley) are cautious but are using AI to expand their capacity and project pipeline:
    "With the same number of people, we can do a lot more." — Vasant Dhar [21:08]
  • Trust remains central: human verification for high-risk outcomes still necessary.

4. Strategic Posture: Look Before You Leap

[24:53–27:28]

• Speed vs. deliberation:

  • "In this case, 'look before you leap' is probably more important than 'he who hesitates is lost.'”
  • Tech is moving so fast that waiting a little may yield advantages (e.g., better capabilities, fewer false starts).
  • BUT: Deliberate thinking is not an excuse for endless inaction.

• On FOMO vs. readiness:

  • "I don't think you're going to get left behind if you wait a week or a month...you get more for your buck the later you start." — Vasant Dhar [25:51]
  • However, experimentation is vital—leaders must understand their data and ask informed questions to assess real readiness.

5. The Imperative for Business Leaders: Think in AI

[29:02–32:09]

• Engagement at all levels required:
  • "Leaders are very far removed from the details of their data...they need to get on board with understanding the quality of what they have." — Vasant Dhar [29:14]
  • Get started, even with “slow thinking.”
• ‘Thinking in AI’—shifting one’s reflex to include AI as a brainstorming, problem-solving partner (not just a technical curiosity).

6. AI History & the “Common Sense + Expertise” Breakthrough

[32:09–43:10]

• The journey from the 1970s expert systems to today’s LLMs:
  • "Our ambitions then were for a machine that could write software…sound familiar?" — Vasant Dhar [32:50]
  • The real leap: LLMs dissolve the boundary between common sense reasoning and domain knowledge.
  • LLMs are “generalists,” capable of mixing expertise and common sense—a major leap beyond prior generations.
  • Anecdote: Predicting the next word (as in Gmail) forced machines to model general world knowledge, leading to today’s GPTs.

7. Is Real Artificial Intelligence Here?

[37:30–43:10]

• Some industry experts say, “we don’t have true AI yet.” Dhar is skeptical:
  • "AI is the only technology designed without a purpose in mind. The goalpost keeps moving…Every time a machine does something, we move it." — Vasant Dhar [38:52]
  • LLMs aren’t human, but they’re intelligent in functional ways.
  • "Why should we expect it to be like us in every way? ... I think we have achieved artificial intelligence to a large degree." — Vasant Dhar [40:14]
  • AI’s status as a general-purpose technology (like electricity): over-delivering after decades of over-promising.

8. The Coming Bifurcation: Amplified vs. Disempowered

[43:55–49:16]

• AI will split users into two groups:

  1. Amplified/Empowered: People with deep domain knowledge who use AI to learn more, ask better questions, and amplify impact.
  2. Disempowered/Dependent: Those who use AI as a crutch (“give me the answer”), risking cognitive decline and loss of agency.

• "The more you know, the more AI can teach you...it's important to be able to ask the right question and gauge the response." — Vasant Dhar [45:15]

• Everyone must ‘opt in’ and learn to think with machines, not compete against them. There’s no opting out.

  • "You don’t have a choice here...It’s not AI vs. humans, but humans thinking with machines." — Chris Daigle [47:26]

• Practical impact: AI amplifies productivity expectations—workers with AI can (and will be asked to) do much more, faster.

9. Practical Action for Business Leaders

[49:25–50:30]

• Key encouragement: Business owners are domain experts—AI is their amplifier if they engage it thoughtfully and ask rich questions.
• Advice: Spend at least 10 minutes daily using AI for your business; breakthroughs will follow.

10. Brave New World—Further Resources

[50:30–54:20]

• Dhar’s Podcast: Brave New World
  • Covers wide-ranging AI impact (business, philosophy, science).
• Newsletter: Substack, Brave New World (https://vasantdhar.substack.com)
• Column: Psychology Today
• Recent & Upcoming Work: Focus on the concept of “fiction machines”—AI doesn’t hallucinate, it “confabulates,” yet somehow produces surprisingly truthful outputs much of the time.

Notable Quotes & Memorable Moments

• "[AI] is a pivotal technology. It’s here to stay…How is it going to transform my business?" — Vasant Dhar [00:34, 18:50]
• "If a machine is never wrong, then you should absolutely trust it…If it’s wrong, but the cost is low, maybe you still trust it." — Vasant Dhar [04:23]
• "You can't just hire talent off the street…You should be skeptical." — Vasant Dhar [10:28]
• "The MIT study showed that 95% of projects failed, not because of the AI, but the wrong problem was chosen. That’s a leadership issue." — Vasant Dhar [11:38]
• "We’re seeing real movement, higher ambitions…we can do more than we used to because this is a real productivity enhancer." — Vasant Dhar [23:44]
• "Look before you leap is probably more important than 'he who hesitates is lost.'" — Vasant Dhar [25:10]
• "Leaders don’t realize they can’t get there from here…until they actually look at their data." — Vasant Dhar [29:14]
• "The difference between modern AI and before is that the boundary between expertise and common sense has broken down." — Vasant Dhar [34:48]
• "I think we have achieved artificial intelligence to a large degree…AI has gone from application to general purpose technology." — Vasant Dhar [40:14]
• "The more you know, the more AI can teach you…It’s the rich get richer phenomenon." — Vasant Dhar [45:01]
• "You don’t have a choice…The future is not right, AI vs. humans, but humans thinking with machines." — Chris Daigle [47:28]
• "Software is the killer app of generative AI. How much can we trust code written by machines?" — Vasant Dhar [22:05]
• "Everything the machine generates is a confabulation…I'm amazed at how often these confabulations turn out to be truthful and correct." — Vasant Dhar [53:00]

Timestamps for Important Segments

• Map of Trust/Trust Framework Intro: [03:51–09:07]
• Top Barriers to AI Adoption: [09:07–16:59]
• Trends/Fear vs. Greed: [18:50–24:03]
• Look Before You Leap: [24:53–27:28]
• The Imperative for Action & Data Readiness: [29:02–32:09]
• History of AI & General Intelligence Paradigm: [32:09–43:10]
• Are We at Real AI?: [37:30–43:10]
• Bifurcation: Amplified vs. Disempowered Citizens: [43:55–49:16]
• Action Advice for Business Leaders: [49:25–50:30]
• Dhar’s Podcast/Newsletter: [50:30–54:20]

Flow & Tone

The conversation is accessible yet rigorous, blending business pragmatism with deep academic insight. Dhar is clear-eyed, evidence-driven, occasionally wry, and always encouraging critical thinking. Chris Daigle keeps the audience anchored in actionable takeaways and reiterates that every leader can and must join this “thinking with machines” revolution.

Further Exploration

• Thinking with Machines: The Brave New World of AI (Book by Vasant Dhar)
• Brave New World Podcast — Interviews on the intersections of technology, business, and humanity
• Brave New World Newsletter (Substack)
• Psychology Today Column by Vasant Dhar

Key Message:
Trust, but always verify. Be deliberate in how you choose AI projects, know your data, apply human judgment—and embrace the future by learning to think with machines, not merely about them.