In 2025, Stanford University launched Stanford Open Minds, a multi-city tour across the United States. The most recent installment, held in September 2025 in New York City, convened some of Stanford’s leading researchers in artificial intelligence, economics, and business strategy to have discussions on “how to navigate the future with open minds.” Here are my three takeaways from the event.
1) An Inflection Point in Education
A hundred years from now, history books may mark this decade as the time when education fundamentally changed.
Language models already perform at the top 10% or above human levels on standardized exams, from the SAT to bar exam and professional certifications. If machines can clear the very benchmarks that have shaped curriculum and assessment for generations, the premise of spending twenty years training youths to do the same no longer holds. The real question is how we build a new framework for human capital development that works in tandem with these advances.
The path forward is not yet clear. Universities are only beginning to wrestle with this question. But the purpose of education will not be to compete with AI. It will be to teach the next generation how to direct these tools toward solving the problems only humans can define.
2) Rethink Work As 100 Tasks
Sixty percent of today’s jobs did not exist 80 years ago. Work has always evolved, and most of what we do today is relatively new.
When we talk about AI and jobs, the first word that often comes to mind is “replacement.” Will a role disappear or survive? That binary framing may be missing the point.
A recurring theme at Stanford Open Minds was the notion that every job is really a portfolio of tasks. In this portfolio, some tasks are recurring and rules-based, others require judgment, creativity and context that only a human in the moment can execute.
For example, a research analyst carries out hundreds of tasks each day. AI agents can handle market research and data pulls, but only the human operator can go out and have coffee with an industry veteran to dig into the nuances of operating in a niche vertical.
So then the right mental model is not a complete replacement of a role but reshaping of individual tasks within the role: automate some tasks, augment others and leave the most human-intensive tasks untouched.
3) Closing the Last Mile of AI
In the coming decade, AI will drive down the cost of software development to the point where software itself qualifies as a general-purpose technology, a foundational capability like electricity.
This change will put advanced tools within reach of small restaurants, local service providers, bootstrapped startups, and regional banks—organizations that until now could not access the capabilities once reserved for Fortune 500 companies.
If generative models make software affordable and faster to build, the role of developers will remain critical, but so will a new class of AI implementation managers. These are the people who close the ‘last mile of adoption’. They are the operators who identify the purpose for the technology, forge the right business partnerships, and ensure it is actually used in organizations.
Even the most elegant AI system is useless if an organization cannot procure it, integrate it, or train its people to use it effectively. Today, many small and mid-sized firms still rely on paper records, outdated systems, or disjointed workflows. The same is true for many government organizations that continue to run on legacy systems built decades ago. These systems carry fixed costs, entrenched data structures, and the inertia of years of institutional habit.
As software and AI become more accessible, the real challenge will not be the tools themselves, but implementation. The people who can make a difference will be the ones who can bring technology into existing organizations: the entrepreneurs, Chief AI Officers, AI implementation managers, and product managers. They will set the strategy, install the systems, and drive adoption.
The true differentiator will be those who can translate AI-enabled builds into solutions that are procurement-ready, accepted by decision makers, and securely integrated.
