In 1890, electricity already existed. Factories could connect to the grid. Engineers understood how the dynamo worked. And yet, industrial productivity took four more decades to take off.
Today, in 2026, many business leaders are living a modern version of that same story. They have access to language models, AI copilots, and BI dashboards. And yet, the results aren’t materializing. Decisions are still slow. Data is still opaque. The organization is still running the same way it always has.
This isn’t a technology problem. It’s an adoption problem.
The Story Paul David Told in 1990
Economist Paul David published a seminal paper in 1990: “The Dynamo and the Computer.” His question was simple: why did electrification — one of the greatest technological revolutions in history — fail to translate into productivity gains for decades?
The answer he found was uncomfortable: companies adopted the new technology without changing anything that made the old technology work.
When electricity arrived, factory owners simply replaced the central steam engine with a central electric motor. They kept the same multi-story vertical architecture, the same drive shafts, the same mechanical belts. They put new technology into an old framework.
The real productivity leap came when someone had the radically different idea of giving each machine its own motor. That made it possible to redesign factories horizontally — optimizing material flow, eliminating bottlenecks, simplifying operations. The process changed, not just the tool.
That took forty years.
The Same Mistake, Repeating Today
Look at what’s happening in most companies with AI right now.
A Copilot license gets purchased so analysts can write code faster. A chatbot gets integrated into the call center. A language model drafts emails. All of that is fine. But it’s exactly the same as plugging a new electric motor into the old drive shaft: new technology, same process.
The result is what David would call the plug-and-play mistake: you adopt the tool but don’t redesign the workflow. AI accelerates individual steps, but the process remains the same process. And the process is the problem.
How many decisions in your company still depend on an analyst who takes three days to build a spreadsheet? How many meetings happen to “review the numbers” that should be available in real time? How many strategic projects stall because someone needs a report that doesn’t exist yet?
AI doesn’t fix that if the process stays the same.
What Actually Needs to Change
The lesson of the dynamo isn’t that you need to wait. It’s that the real value of a general-purpose technology only emerges when you redesign the process from scratch around it.
For AI applied to data — where the decisions that move the business actually get made — that means moving from a model where data is an input for reports, to a model where data is the living infrastructure on which the company operates.
It’s not about “having a dashboard.” It’s about the commercial leader being able to understand why sales dropped in a region in minutes, not days. It’s not about “doing analysis.” It’s about systems detecting anomalies before they become problems, alerting, and acting.
That difference — between AI that assists tasks and AI that transforms processes — is exactly the gap that will separate the companies that lead the next decade from the ones that keep playing catch-up.
The New Bottleneck: Data
There’s an irony in all of this. The companies that invest most in AI are often the ones that benefit least from it, because they run into the same wall: their data isn’t ready for AI to work on.
Data is scattered across silos. There are three different versions of the same metric depending on who you ask. The ERP doesn’t talk to the CRM. Reports are built by hand every month. And nobody knows which is the source of truth.
In that context, the most sophisticated AI in the world can’t do anything useful. It’s like having the best electric motor available plugged into a factory where nobody knows where the wiring is.
The real bottleneck isn’t the language model. It’s the data infrastructure.
The Factory Redesigned for the AI Era
What the most advanced companies are building today — and what will define who wins over the next several years — is what we might call a modern data factory: an infrastructure where data flows automatically, is always up to date, is trustworthy, and is available for consumption by AI, by people, or by systems in real time.
This isn’t a two-year IT project. It doesn’t require a team of 20 data engineers. That was the old model.
The new model combines specialized AI agents that build and maintain that infrastructure autonomously — cutting implementation time from months to days, and maintenance cost to a fraction of what it used to require.
The result is concrete: a business leader can explore their operational data, detect patterns, anticipate problems, and make decisions — without waiting for a report, without asking IT for anything, without depending on an analyst juggling ten other things at the same time.
That’s what redesigning the process looks like, not just changing the tool.
Strategic Patience, Not Passive Waiting
Paul David ended his paper with an idea worth repeating: the productivity lag isn’t a signal that the technology is failing. It’s a signal that real transformation takes time — but more than anything, it takes intent.
The companies that won from electrification weren’t the first to buy an electric motor. They were the ones with the vision to redesign the entire factory.
The companies that will win from AI aren’t necessarily the ones spending the most on model licenses. They’re the ones that understand the value isn’t in the tool — it’s in how the entire workflow gets reorganized around it.
And at the center of that workflow — at the heart of every strategic decision — there is always data. Data that must be reliable, current, and actionable.
That’s the factory worth building today.