AI is rapidly improving what people can produce. But is it also weakening the capabilities that make people valuable in the first place?
In this episode of Partnering Leadership, Mahan Tavakoli speaks with Eric So, Stanford-trained behavioral researcher, longtime MIT professor, and author of The Collision: What AI Does to Us. Eric explores a question that sits underneath much of today’s excitement about artificial intelligence: not simply what AI can do for us, but what sustained reliance on AI may be doing to how we think, learn, judge, and work.
Eric explains why AI represents a more consequential shift than previous cognitive tools. Calculators automated pieces of thinking. AI increasingly offers to perform the entire chain, from interpreting a problem to generating an answer. That creates a new risk: people can produce polished work while bypassing the struggle that develops understanding and mastery. The result can be better output without deeper capability.
The conversation then turns to what this means inside organizations. Eric introduces the idea of “signature thinking skills,” the forms of judgment and expertise that distinguish one professional from another. He argues that organizations should be careful not to incentivize AI use for its own sake, because doing so can encourage employees to outsource precisely the capabilities they most need to preserve.
Eric also challenges the common assumption that strong human judgment combined with strong AI will automatically produce better results. His “dream team fallacy” highlights how anchoring, complacency, and overreliance can make human-AI collaboration worse than expected. That leads to a counterintuitive design principle: sometimes organizations need to preserve friction so people continue to think, question, and remain engaged rather than becoming passive recipients of machine-generated recommendations.
Finally, the conversation examines what happens when cognitive outsourcing moves from the individual to the organization. Eric raises concerns about lost apprenticeship, weakened institutional memory, reduced knowledge transfer, and the erosion of trust when people increasingly turn to AI instead of one another. At the same time, he makes clear that the goal is not to reject AI. The opportunity is to use it in ways that expand human capability, reinvest the time it saves into higher-value thinking, and make people better off, not merely faster.
Actionable Takeaways
- You’ll learn why Eric So believes AI is fundamentally different from calculators, search engines, and other tools because it can increasingly take over the full chain of thinking, not just one step in it.
- Hear how to distinguish between using AI to amplify your work and outsourcing the “signature thinking skills” that actually define your professional value.
- You’ll learn why polished AI-generated output can create a dangerous illusion of mastery, and why being able to explain or defend an idea may reveal more than the quality of the artifact itself.
- Hear how incentive systems can unintentionally push employees to overuse AI, and why rewarding outcomes rather than raw AI activity may lead to better organizational behavior.
- You’ll learn why “great human + great AI” does not automatically produce better performance, and how anchoring and complacency can undermine the value of human judgment.
- Hear why introducing deliberate friction into AI interactions can sometimes improve reasoning, retention, and the quality of human-machine collaboration.
- You’ll learn how short-term efficiency gains from reducing junior roles may create longer-term risks for expertise, succession, and organizational memory.
- Hear how to think about AI as a cognitive trainer rather than simply an answering machine, and why reinvesting the time AI saves may be more important than using it to produce more of the same.
Connect with Eric So
Eric So Website
Eric So LinkedIn
Connect with Mahan Tavakoli:
Mahan Tavakoli Website
Mahan Tavakoli on LinkedIn
Partnering Leadership Website
