The ice man cometh.
Everyone is talking about AI and SaaS. I keep thinking about ice.
Before refrigeration, there was an enormous industry dedicated to harvesting, storing, transporting, and delivering ice. Thousands of workers. Specialized equipment. Distribution networks. Real businesses creating real value.
Then refrigeration arrived.
People often say the ice industry died. It did not. Demand for cold exploded. What died was the need to buy ice.
Customers never really wanted ice. They wanted cold.
That is the question sitting underneath software right now. Are today's SaaS companies selling software, or are they selling outcomes?
A reservation system is not really software. It is court reservations. A CRM is not really software. It is customer management. An accounting system is not really software. It is financial control.
If AI changes how those outcomes are delivered, we may be seeing the beginning of a transition from SaaS to something else entirely.
I think of this first as the CaaS lesson: customers do not buy the product. They buy the result.
Kodak's real mistake.
People often cite Kodak as a company that missed digital photography. That is not quite right. Kodak understood digital photography. Kodak actually helped invent it.
The problem was deeper. Kodak was organized around film: factories, chemicals, distribution, processing labs, and revenue streams.
Customers did not want film. They wanted photographs.
When digital became good enough, the value shifted away from the infrastructure that had made Kodak successful. I wonder if some SaaS companies face a similar challenge.
Their assets are not factories. They are massive code bases, implementation teams, training departments, support organizations, and legacy integrations.
Those are valuable assets if software remains difficult and expensive to create. But what if AI changes that equation?
History suggests that industries are often disrupted not because they fail to see the future, but because they are heavily invested in the present.
Sea Oaks and the $100,000 question.
At my tennis club, Sea Oaks, there has been discussion about replacing parts of our club management software. Not replacing it all at once. One piece at a time.
- Reservations first.
- Then reporting.
- Then communication.
- Then other functions.
Eventually someone asks, "Why are we still paying for the old system?"
The interesting thing is that nobody planned a revolution. Each step is individually rational. Historically, a small organization like ours could never have justified custom software development. Today, with AI-assisted development, that assumption is changing.
The economics are changing from, "Can we afford custom software?" to "Can we afford not to have software that exactly fits our needs?"
That feels like a very different world.
What if AI creates more programmers?
Most predictions assume AI reduces programming jobs. Maybe. But history suggests another possibility.
In 1950, almost no company had programmers. No CIOs. No IT managers. No network administrators. No database specialists.
Computers became dramatically easier to use. The result was not fewer technology jobs. The result was millions more.
Technology became distributed throughout the economy. Perhaps AI does something similar.
Maybe the future is not fewer programmers. Maybe it is software expertise embedded in restaurants, medical practices, tennis clubs, nonprofits, schools, and local businesses.
The job changes. The distribution changes. But the number of people creating and managing software may actually grow.
Open source changes everything.
There is another difference between software and most historical industries. Programmers have spent decades building a culture of sharing.
Open source, GitHub, Stack Overflow, frameworks, and libraries mean that a programmer today stands on top of millions of hours of donated effort.
AI amplifies this effect. One person can increasingly leverage the accumulated knowledge of an entire profession.
That changes the economics. The question may no longer be, "How many programmers can we afford?"
It may become, "Do we have someone who understands our business well enough to direct the software?"
Domain knowledge becomes the scarce resource. Not syntax.
From SaaS to OaaS.
My current hypothesis is that the future may not be SaaS. It may not even be CaaS. It may be OaaS: Outcomes as a Service.
Nobody buys accounting software because they love accounting software. Nobody buys reservation software because they love reservation software.
They buy outcomes. They want accurate books, filled courts, managed inventory, and better customer relationships.
If AI continues to lower the cost of creating software, businesses may increasingly own highly customized systems built around their unique operations. The software becomes invisible, just as refrigeration made ice invisible.
That question may determine the next Kodak, or the next refrigeration company.
The old ice companies thought they were selling ice. The durable business was cold. Some software companies may be making the same mistake. They may think they are selling software when the customer is really buying the outcome.
Back to writing