The Incumbent's Calm

Executive summary: Most companies are not ignoring AI. They are welcoming it in the most dangerous way, as a sustaining upgrade to the business they already run, when it is a disruptive technology that rewards a different way of working. A sustaining technology makes your current product better for your current customers, and AI used that way, the copilots, the agents, the automations, the efficiency gains, is real and worth doing. It is also not a response to the disruption. The companies pulling ahead are not the ones with the most copilots. They are the small share building a fresh curve while the old one still pays, and the performance gap between them and everyone else is already wide and widening. This piece is about telling the two apart, and about the nerve to jump to the new curve while the old one still feels fine.

This spring, several technology luminaries walked onto commencement stages to tell graduates that something historic had arrived. Several were booed. At the University of Central Florida on May 8, a development executive called AI the next industrial revolution and the crowd turned on her. She paused and asked to finish. A week later, when Eric Schmidt compared AI to the dawn of computing at the University of Arizona, the boos rose as he made the comparison. The motives in any one crowd are mixed and not mine to read. The pattern across the season is the part worth holding.

Set that against a number. In the 2025 Stanford AI Index, 83 percent of people in China saw AI as more beneficial than harmful. In the United States, 39 percent. The richer and more established the country, the cooler the enthusiasm. The 2026 edition makes the divide sharper at home: American experts and the public split by roughly fifty points on whether AI will help or hurt jobs. The people building the technology and the people who will be sorted by it are reading two different futures.

So the people with the most to gain sound the most nervous, and some of the graduates who will live longest with the technology boo the optimists. The contradiction resolves once you see what the calm are actually getting wrong. They are not ignoring AI. They have misread what kind of technology it is.

Sustaining or disruptive

Clayton Christensen drew the line almost thirty years ago, and for the executive it is the only distinction that matters. A sustaining technology makes an existing product better along the lines current customers already value. The next phone with the sharper camera. It is welcome, it is easy to fund, and it keeps you climbing the curve you are already on. A disruptive technology arrives differently. It is a new way of doing the thing, worse at first on what your best customers care about, but cheaper or simpler or more accessible, and easy to wave off as a toy. Then it improves, and it becomes the substitute that takes the market. The personal computer did that to the mainframe. One-hour photo labs did it to instant film. The disruptor almost never wins by being better on day one. It wins by being good enough on a new dimension and then climbing.

Here is the deception, and it is doing real damage right now. AI is wearing the sustaining costume. It shows up as copilots inside the tools you already own, agents that run a task end to end, automations that trim a workflow, a faster draft, a ticket closed quicker. All of it is genuinely useful, and I would not talk any operator out of the efficiency. All of it is also sustaining. It makes the old curve a little taller. Building a new way of doing the work is a different act entirely, and the distance between those two responses is the distance between the companies that will compound and the companies that will fall behind while feeling busy and modern.

Run a simple test in your next meeting. Ask what the AI in front of you actually changes. If it makes your current product better for the customers you already serve, it is sustaining. Buy it, and do not confuse it for strategy. If instead it lets someone serve those customers in a way that makes your current product unnecessary, it is disruptive, and no amount of bolting it onto your old process will save that process. Most AI spending today quietly fails this test, and almost no one says so out loud.

That is the trap in one sentence. Layering AI onto how you already work feels like progress and quietly keeps you on the curve that is going to flatten.

And do not let the copilot framing shrink what is arriving. The disruption is not a smarter assistant. It is the commoditization of intelligence itself. For the entire history of modern business, the scarce and expensive input was human cognition: the analyst, the coder, the underwriter, the associate in the back room. That input is becoming cheap and abundant, and when the price of intelligence collapses you do not get a better version of the old company. You get a different shape of company, with different work, different headcount, and different economics. Some argue the threshold is already behind us, that what we politely call assistants are early general intelligence. You do not have to settle that debate to act on it. The rudder has already shifted, and this is the infancy, not the peak.

Every curve has a top

Richard Foster gave the shape its name. Every technology climbs an S-curve and flattens at a plateau. The incumbent sits at the top collecting fat margins, which is exactly where the next curve, low and toy-like, is easiest to dismiss. Leaving a plateau you are still profiting from, to climb a curve that underperforms it today, feels like malpractice right up until it is the only thing that would have saved you.

The cruel part is that the plateau is comfortable. The numbers are good, the team is proud, the board is pleased. Nothing on the dashboard warns you that the curve beneath you has stopped climbing. By the time the slowdown shows up in the results, the cheap years to act are already spent.

The cleanest naming of that arc comes from the attorney Matt Brown, who set out four stages of the entrepreneurial life cycle. Wonder is the start, the what-if, when the founder is sure there must be a better way. Blunder is the baptism by fire, the real-world education that arrives once the plunge is taken and the mistakes begin. Thunder is when the business strikes a rhythm, growth takes hold, and confidence grows with it. Plunder is the arrival, the I-have-made-it stage, when the money comes easy and the only question left is what next: reinvest, reinvent, sell, or hand it on.

Brown named four stages. The companies in this piece teach a fifth, and it does not rhyme by accident. Under is what happens when a business treats plunder as a destination instead of a launch point. The comfort of plunder is the incumbent's calm. Sit in it and harvest it, and you slide toward under while the next curve climbs out of reach. The discipline the winners force on everyone else is to jump in thunder, while the current business is still rising and the leap looks unnecessary, spending the strength of one curve to start the next. The copilot keeps you climbing the curve you are already on. The jump is the other thing: a new curve, a new way of doing the work, begun while the old one still pays.

The gap is already open

This is not a someday problem, and the numbers say so. Boston Consulting Group's 2025 study of how companies are actually faring with AI sorts them into three groups. About 5 percent are what BCG calls future-built, redesigning how they work around AI and pulling away. Against the laggards they show 1.7 times the revenue growth, 1.6 times the operating margin, and 3.6 times the three-year shareholder return. Roughly a third are scaling. About 60 percent are getting little for their spend. McKinsey's 2025 read agrees from the other side: only about 6 percent are capturing real bottom-line impact, and what sets them apart is that they redesign the work rather than bolt AI onto it. Thomson Reuters found the same split from another angle: firms with a real AI strategy are about twice as likely to see revenue growth from it as the ones improvising. The finding is consistent across every serious study this year. The money follows the companies that change the work, not the ones that decorate it.

Sit with the uncomfortable part. Most of that 60 percent are not standing still. They are buying AI, running pilots, sending the reskilling memo. They are doing AI in sustaining mode, on the old curve, which is exactly why they stay laggards on value. Adoption is not advantage. The advantage goes to the few who change the work itself. The diffusion curve Everett Rogers described decades ago still holds: a thin band of innovators and early adopters moves first, the majority follows slowly, the laggards last. What is different now is the prize for being in that front band, and the speed at which the door is closing. The front band is about 5 percent, and the gap behind it widens every quarter.

I have stood at the doorway

I trust this pattern because I have stood at the doorway of every version of it. I studied computer science in the early 1980s and watched the mainframe give way to the personal computer, then the internet, then the move to cloud and mobile, with quantum waiting its turn. Each wave made the last winner's advantage close to irrelevant. Nearly every time, the incumbent that lost had the new technology in the building and chose to treat it as a feature.

Four companies, one mistake

Kodak built the first digital camera in 1975 and went on to hold more than a thousand digital patents. It saw the future early enough to own it on paper. What it would not do was let digital replace film, the business that paid for everything, so it treated the new curve as an accessory to the old one. Kodak filed for bankruptcy in January 2012.

Google invented the architecture behind today's chatbots and had a capable system of its own. It hesitated to ship, because a machine that answers your question completely does not show the ads that make up most of its revenue. When ChatGPT arrived at the end of 2022, Google declared an internal code red. The technology was in the building; the old curve fought its release. The tempo is the warning. By the end of 2025, the upstart that had beaten Google declared a code red of its own as Google surged back. Three years was enough to swap the roles.

Chegg sold homework help on a subscription until free AI answers did the same work for nothing, and its customers simply walked onto the new curve. In the first quarter of 2025 its revenue fell about a third and its subscriber base about a third, and across the year it cut more than half its staff. No better product beat Chegg. A good-enough free one did.

DeepSeek ran the same play at the scale of a nation. In January 2025 a Chinese lab released a capable model that was free and, by its own account, far cheaper to build than its American rivals, and in a single day it took close to six hundred billion dollars off Nvidia, the largest one-day loss in the history of the United States market. The same country now installs more industrial robots each year than the rest of the world combined. Cheaper, good enough, and running on a new curve, carried by a country rather than a startup.

Build the new skin

So the board question is not whether you are using AI. Almost everyone is. The question is which curve your AI is on. Is it making your current way of working faster, or building a new way of working? If it is the first, you are climbing a plateau and calling it strategy. Pick one bet that is unmistakably the new curve, a different way of delivering what you deliver. Put someone senior on it, measure them on building the new thing rather than improving the old one, and fund it now, in thunder, while the old curve still pays for it.

A new-curve bet is not a bigger copilot-and-agent budget. It looks like a small team given permission to deliver what you deliver in a way that would embarrass your current operation, with its own number to hit and cover from the top to chase it. If the only AI plan on the table is a tool rollout and a training schedule, you do not have a strategy yet. You have a faster horse.

The teacher in me wants to leave you with the oldest version of the lesson. You do not pour new wine into old wineskins. The new wine ferments, the rigid old skins burst, and you lose the wine and the skins both. AI is new wine. Your existing processes are the old skins. Layer the one onto the other and you lose both. The work is to build the new skin.

The graduates who booed may have the timing wrong. They do not have the direction wrong. And if this is what the technology does to companies, the harder question is what it does to work, and to the people who do it. That is where this goes next.

References

Clayton M. Christensen, The Innovator's Dilemma (Harvard Business School Press, 1997); Clayton M. Christensen, Michael E. Raynor, and Rory McDonald, "What Is Disruptive Innovation?" Harvard Business Review, December 2015. The sustaining and disruptive distinction and the personal-computer-over-mainframe example follow Christensen; the one-hour-photo-over-instant-film case is used as illustration.

Richard N. Foster, Innovation: The Attacker's Advantage (Summit Books, 1986).

Everett M. Rogers, Diffusion of Innovations (Free Press, 1962).

The four-stage entrepreneurial life cycle, Wonder, Blunder, Thunder, Plunder, is from Matt Brown, "The Lifecycle of the Entrepreneurial Business: Wonder, Blunder, Thunder, Plunder," Family Enterprise USA, February 10, 2022. The fifth stage, under, is the author's own addition.

Stanford Institute for Human-Centered AI, Artificial Intelligence Index Report 2025 (Ipsos public-opinion data, 2024) and 2026 (expert-versus-public divide).

NBC News and Fox Business on Eric Schmidt at the University of Arizona commencement, May 2026; 404 Media on the University of Central Florida commencement, May 8, 2026.

Boston Consulting Group, "The Widening AI Value Gap" (Build for the Future 2025), September 2025: about 5 percent future-built firms versus roughly 60 percent laggards, with 1.7 times revenue growth, 1.6 times EBIT margin, and 3.6 times three-year total shareholder return.

McKinsey and Company, "The State of AI in 2025": about 6 percent of firms capture significant bottom-line impact, distinguished by redesigning workflows rather than adding AI to existing ones.

Thomson Reuters, "The AI Adoption Reality Check," June 2025: firms with a defined AI strategy are roughly twice as likely to see AI-driven revenue growth.

Computer History Museum and IEEE Spectrum on Kodak's 1975 digital camera and digital-imaging patents; Kodak Chapter 11 filing, January 2012.

Vaswani et al., "Attention Is All You Need" (Google, 2017); The New York Times and CNBC on Google's code red, December 2022; Fortune, The Information, and The Wall Street Journal on OpenAI's code red, December 2025.

Chegg Inc. SEC filings, Forms 8-K, first through third quarter 2025; Chegg Inc. v. Google LLC, U.S. District Court for the District of Columbia (filed February 24, 2025).

DeepSeek R1 release, January 2025; CNBC and Bloomberg on the Nvidia single-day loss of approximately 589 billion dollars, January 27, 2025; International Federation of Robotics, World Robotics 2025, on China's industrial robot installations.