Not That GPT: AI as a Generally Productive Technology

Every time I start a talk with the phrase "let me tell you about GPT," I watch the same thing happen in the room. Heads go down, phones come out, or eyebrows raise with the polite-but-skeptical look of someone who is about to hear a demo of a chatbot they have already tried.

That is not the talk I am giving.

When I say GPT, I mean Generally Productive Technology. And the reason that distinction matters is that it changes everything about how you think about AI, what it demands of you as a leader, and how much time you have to respond.

What Is a Generally Productive Technology

Economists use the term "General Purpose Technology" for a category of innovations that produce macro-economic effects so broad that they restructure entire economies rather than improving specific sectors. The canonical examples: the steam engine, electricity, the railroad, the internet.

GPTs share three characteristics. First, they are applicable across many domains — they are not solutions to specific problems but enablers of new solutions across virtually every economic activity. Second, they produce continuing innovation — the initial adoption opens new possibilities that were not visible before the technology existed. Third, they create network effects — the more broadly they are adopted, the more productive they become for all participants.

AI is a Generally Productive Technology. The deliberate play on the OpenAI name is the point: the company name has made "GPT" synonymous with a specific product category, when the more accurate framing is a macro-economic category that will reshape the labor market, the capital market, and the competitive landscape in every sector.

Why the Reframe Changes the Conversation

When you think of AI as a product category — a set of tools with specific features and use cases — the questions you ask are: what can it do? How much does it cost? What are the risks? These are legitimate questions, but they are downstream of the more important ones.

When you think of AI as a Generally Productive Technology — a macro-economic force comparable to electricity — the questions change: how does the availability of this technology restructure the economics of our industry? What business models that were previously infeasible become viable? What competitive positions that were previously defensible become vulnerable? What skills and capabilities will be scarce vs. abundant in five years?

These are the questions executives asked about the internet in the mid-1990s. The organizations that answered them correctly in 1996 built Amazon. The organizations that said "the internet is a new customer service channel" were Circuit City and Borders.

The Electricity Analogy

Electricity is the analogy I find most useful, and it is worth spending time on.

Before widespread electrification, industrial production ran on mechanical power: water wheels, steam engines, and the complex systems of shafts, pulleys, and belts that distributed that power through a factory. The factory layout was dictated by the power distribution infrastructure — machines clustered near the power source, organized in fixed configurations determined by the physical transmission of mechanical force.

Electricity did not simply make factories more efficient. It eliminated the constraint that had organized factories for decades. With electric motors, each machine could have its own power source. The factory layout was no longer dictated by power distribution. The whole concept of how a factory was organized had to be rethought.

The organizations that thrived in the electrification era were not the ones that found the most efficient way to add electric motors to their existing shaft-and-pulley systems. They were the ones that recognized that electrification made the existing factory organization obsolete and redesigned from scratch.

AI is doing the same thing to knowledge work. It is not making existing knowledge work processes more efficient. It is eliminating the constraint — human execution capacity — that organized those processes for decades.

What This Means for Timing

GPTs do not wait for organizations to be ready. The electrification analogy is instructive here too: the companies that did not adopt electric power by a certain point were not merely slower — they were operating at a structural cost disadvantage that made them uncompetitive regardless of their other capabilities.

The capability data is clear: AI models are now performing at or above human expert level on specific benchmarks (PhD-level science reasoning, professional software development, legal analysis, medical diagnosis). The cost data is equally clear: the price per unit of AI-generated output has fallen 9–900x per year depending on the model tier.

The combination of rising capability and falling cost is the economic condition that produces rapid adoption across markets. It does not produce uniform adoption — the Rogers curve will play out, with Innovators and Early Adopters capturing most of the advantage before Early Majority adoption commoditizes the gains. But the window for Early Adopter positioning is measured in months to low single-digit years, not decades.

The practical question is not "should we use AI?" That is the wrong question, and the organizations asking it are already at risk. The question is: "How do we position ourselves in this transition such that we capture the advantage rather than bear the disruption?"

Starting the Right Conversation

The reframe I offer in every executive workshop: stop asking "what can AI do for us?" and start asking "how does AI change the economics of what we do?"

The first question produces a list of tools. The second produces a strategy.

Part of the Thought Leadership series — Thread 3: AI & Machine Learning. Related: [[T22-labor-compression]], [[T28-capability-cost-inflection]], [[X04-diffusion-curve-meets-ai-adoption]]