Exponential Kodak

In 1975, Kodak invented digital photography. They buried it. In 2000, Blockbuster could have bought Netflix for $50 million. They laughed. The pattern keeps repeating, and right now, it's happening with generative AI. Only faster: exponentially, then instantaneously.

You know the Kodak story. In 1975, their own engineer built the first digital camera. Executives called it "cute" and told him not to talk about it. What's often missed: Kodak ran internal studies predicting digital would replace film within 10 to 20 years. They had the data. They had the prototypes. They still chose to use digital only to improve film quality rather than disrupt their own business. The breakthrough wasn't missing. The willingness to cannibalise their own cash cow was.

Kodak isn't alone. Xerox PARC built the Alto in the late 1970s: a machine with a mouse, graphical windows, WYSIWYG text, and networking, decades ahead of its time. Headquarters treated it as a research toy. Steve Jobs walked in, absorbed everything, and built the Macintosh. In September 2000, Netflix founders flew to Blockbuster's Dallas headquarters and offered to sell their struggling DVD-by-mail startup for $50 million. Blockbuster CEO John Antioco reportedly laughed off the idea. Within a decade, Blockbuster filed for bankruptcy whilst Netflix dominated home entertainment.

The incumbents didn't lack vision. They had prototypes, data, forecasts, even offers to buy the disruption. They just couldn't stomach the short-term pain of eating their own lunch before someone else did.

I've lived this before

This generative AI moment feels eerily familiar to me. Not because I read about Kodak, but because I experienced the same dynamic personally, about 25 years ago.

Back then, I was struggling with business intelligence automation using iReport, an open-source visual designer for JasperReports. I spent hours — sometimes six or seven — creating a single report that would save me maybe ten minutes of work per week. The maths looked terrible at first.

But I kept going. One report led to a second, a third, a fourth. James Clear calls this the "plateau of latent potential" in Atomic Habits: you put in work that seems to produce nothing, then suddenly everything compounds at once. That's exactly what happened. The more I built, the faster I got. The faster I got, the more I built.

All my colleagues had one or two administrative employees. I had zero. But cost savings weren't even the biggest win. The real payoff was insight. My colleagues hired expensive consultants for budgeting, payroll analysis, and forecasting. I didn't need consultants because I had dashboards with real-time information. I could see problems emerging and intervene immediately. I made better decisions, faster, with actual data instead of gut feelings and quarterly reports.

To be fair, my total headcount was similar to my peers. The difference was where I invested: more caregivers, nurses, and physiotherapists. Same cost, radically different capability.

Clear's point about habits applies equally to organisational capabilities: the results don't arrive linearly. You struggle, struggle, struggle, and then suddenly you're operating at a level others can't match.

Exponentially, then instantaneously

I'm sharing this because I see the same reluctance today with generative AI that I saw with business intelligence tools 25 years ago, and that Kodak showed with digital photography.

The excuse is always the same: "It's too early." "It's not mature enough." "We'll adopt it when it proves itself."

By then, you're Blockbuster watching Netflix eat your lunch.

Ernest Hemingway wrote about bankruptcy: "Gradually, and then suddenly." With AI, I'd revise that: Exponentially, then instantaneously.

The adoption curves are steeper. The capability improvements are faster. The window for catching up is narrower.

If your company isn't experimenting with generative AI right now, if your teams aren't painstakingly struggling with it at 2am to learn what it can and can't do, I give you two to three years. That's the timeline I see for many businesses to simply stop existing.

Most people around me dramatically underestimate how fast and how impactful this change will come. Yes, there will be layoffs. But there will also be displacement, new types of work, and massive opportunities for those who started early enough to compound their learning.

The experiment I'd suggest

Pick one repetitive task in your work. Spend two hours trying to automate or augment it with AI. Document what worked, what failed, what surprised you. Repeat next week. And the week after.

The snowball starts small. It always does.