Blog

Over the last two weeks, I watched first-year MBA students at Tuck react in amazement as I instantly created a song about the excitement of starting business school using AI. Their eyes lit up as Anthropic’s Claude produced interactive tools and simulations, Google’s NotebookLM synthesized complex readings and turned them into a synthetic podcast for the reader, and OpenAI’s ChatGPT solved problems in real time. For them, AI wasn’t a distant concept. It was an immediate, practical tool for creativity and productivity.

One student’s story stood out. Before Tuck, she worked at a global retailer where she was forbidden from using AI. This ban extended not just to consumer tools like ChatGPT, but even to AI features embedded in licensed enterprise software the company was already paying for. Think about that: a firm blocking employees from capabilities built into its own tools.

The contrast with AI-native companies couldn’t be clearer. MBA students are curious but still novices, experimenting at the edges. AI-native startups, meanwhile, are rewiring workflows around intelligent systems. Legacy enterprises, in contrast, are too often stuck in efficiency-focused thinking. They treat AI as a headcount reducer rather than a strategic catalyst. Some are even stifling employee experimentation. That mindset is not just limiting. It risks strategic obsolescence.

The AI Maturity Gap

Recent research from MIT underscores the problem. According to MIT NANDA’s report, The GenAI Divide: State of AI in Business 2025, despite $30–40 billion invested in GenAI, 95% of organizations see no return. Most deployments stop at boosting individual productivity: faster emails, quicker research, smoother presentations. Only 5 percent of enterprise-grade projects reach production.

Here is the divide:

• AI leaders use intelligent systems constantly and assume the rest of the world is catching up
• Legacy enterprises launch endless pilots that rarely scale or produce meaningful returns
• Employees often adopt AI tools informally (over 90% report using them), while only 40% of companies purchase official subscriptions

This “shadow AI economy” reveals a disconnect between individual enthusiasm and organizational transformation, while also revealing a strategy flaw at many legacy companies. 

Efficiency vs. Reinvention: The Amazon and Sears Lesson

We have seen this movie before. When e-commerce emerged, Sears, the retail giant of its era, treated the web as just another catalog channel. Their assumption was that the existing model would hold and their brand reputation and market power were moats around their business.

Amazon took the opposite approach. It built new business architectures: two-sided marketplaces, data-driven logistics, reimagined customer experiences, and eventually cloud computing. Amazon did not digitize Sears. It became something entirely different.

The lesson is clear. The internet was not about efficiency. It was about reinvention. Those who optimized for yesterday’s model lost. Those who reimagined business from first principles thrived.

AI represents the same kind of inflection point. Like the Internet, it will not just make existing businesses faster. It will rewire how they operate. The 5 percent of organizations crossing MIT’s “GenAI Divide” are already building adaptive systems that learn, remember, and evolve. Everyone else is stuck polishing legacy processes, looking for incremental gains. 

Why Incrementalism Is Not Enough

To be fair, efficiency plays are not inherently wrong. For heavily regulated industries such as finance or healthcare, small pilots reduce risk and build trust. But stopping there is a mistake. Incremental adoption does not prepare organizations for competitors designing entirely new models around AI.

Consider what is already happening:

• Walmart uses AI to optimize supply chains and improve in-store experiences
• JPMorgan deploys AI on trading desks and for fraud detection
• Pharma firms are cutting drug development timelines with AI-driven simulations

These are not experiments in efficiency. They are steps toward new operating models. The real risk is mistaking task automation for transformation.

First Principles Thinking as Strategic Advantage

So what should leaders do? Apply first principles thinking: break problems down to fundamentals and rebuild from there. Instead of asking, “How can AI make our operations more efficient?”, ask:

• What is our business truly trying to accomplish?
• If cognitive tasks were nearly free, how would we design differently?
• Which assumptions about value chains, workflows, or structures no longer hold?
• What new customer problems could we solve with unlimited analytical capability?

This mindset forces a reset. Many organizational structures exist because humans have cognitive limits. AI shifts those limits. Entirely new business models are possible when decisions can be made at machine speed and scale.

The MBA Connection

For MBA students, this divide is both a challenge and an opportunity. Business schools teach optimization within existing frameworks. These skills remain important, but they are not sufficient in the AI era. The leaders who will stand out are those who can:

• Think from first principles about business design
• Translate AI-native possibilities into legacy organizations
• Build systems thinking, comfort with ambiguity, and fluency in AI capabilities

In short, you will need to help enterprises imagine new possibilities, not just squeeze more efficiency from old ones.

A Practical Framework: Crossing the GenAI Divide

As a leader, you can use MIT’s research to assess where your organization stands. Here is a simple framework for accomplishing that assessment:

1. Audit your AI approach:

• Are you building custom tools internally (33 percent success) vs. partnering externally (67 percent success)?
• Are you investing mostly in sales and marketing vs. high-ROI back-office operations?
• Are you deploying static tools vs. adaptive systems that learn from feedback?

2. Spot the warning signs:

• Are you on “the wrong side” of AI development: constant prompting, no memory, endless pilots, no production
• Are you on “the right side” of AI development: deep integration into workflows (or developing new workflows entirely), adaptive learning, measurable P&L impact

3. Shift the framing:

• Do not ask: “How can AI make our vendor 20 percent more efficient?”
• Ask: “How can AI eliminate the need for this vendor entirely while improving quality?”

The difference is what separates the 5 percent achieving transformation from the 95 percent stuck in marginal gains.

Final Thought

The MBA students I met last week were both amazed by AI and aware they are just beginning. Legacy enterprises are in the same position. The question is whether leaders will move beyond amazement and incrementalism, or repeat Sears’ mistake and optimize their way into irrelevance.

For AI leaders, the challenge is equally clear. The gap between your AI-native reality and most enterprises is larger than you think. Bridging it is not just good business. It is the only way AI will fulfill its potential to reshape how the world works.

Discussion Question: Which industry do you think is most vulnerable to “Sears thinking” about AI, and what would an “Amazon approach” look like there? 

In case you’re curious, you can listen to the song I created to demonstrate AI capabilities, titled, “The Weight of the Casebooks,” below. The song was created using a free Suno account and is not licensed for commercial use.