The AI Productivity Boom Hiding in Plain Sight

I keep hearing that artificial intelligence (AI) is not delivering on promises of booming productivity.

Economists look at national statistics and see only modest gains. Skeptics conclude that AI is overhyped—another technology long on promise and short on measurable gains.

I heard the same thing back in the early 1990s when a plethora of distinguished experts, many sadly burdened with PhDs, insisted that computers till that point had made no discernible contribution to national productivity.

The problem then was not with the computers; it is not with AI today. The problem is with “national productivity” and all the other silly statistics with which macro economists delude themselves. Essentially, nothing interesting about an economy can be observed on the macro level.

Nations prosper by innovation and determination, which is to say, by their entrepreneurs whose success always comes as a surprise.

Well, here is a surprise:

Midjourney is one of the most widely used AI image platforms in the world, serving millions of users. Its employee head count is “somewhat” smaller, reportedly around 10-20 people, just enough to fit comfortably in a conference room if elbows are kept in check. Meanwhile, revenues are reportedly about $200 million.

That would put revenues at $10 to $15 million per employee.

Midjourney is far from typical. It has some special advantages. It is consumer-facing; it does not need sales teams, nor compliance groups. And it’s small.

So how about OpenAI? It is widely reported to have a revenue run-rate of $1.5–$2 billion (2024 estimates, likely double that today) versus perhaps 1,500 employees.

Revenue per employee exceeds $1 million.

The venture capital community now routinely reports startups reaching tens of millions in revenue on a small fraction of the headcount that would have been typical even five years ago.

Sam Altman has suggested that a one-person billion-dollar company is now conceivable.

Already, surprise is becoming expectation. At Shopify (NASDAQ: SHOP), management has begun to require that before a new human is hired, teams must prove that AI cannot do the same job.

Here’s what’s changed. For decades, the size of organizations has been limited not by ambition, but by coordination. Peter Drucker observed that a manager could effectively supervise only a handful of direct reports, maybe 5-7. That constraint forced companies into pyramiding layers of management simply to move information up and down the hierarchy.

Assuming six direct reports, a company of 1,500 employees will have a minimum of five layers of management, but likely more in practice. As organizations grow, coordination costs grow with them. Meetings multiply. Reporting structures thicken. Decisions slow.

Yet companies like OpenAI are reaching comparable scope with far fewer layers. Software is absorbing part of the coordination burden that hierarchy once carried.

Scale has always carried a hidden tax. Artificial intelligence is cutting tax rates.

It does this not by replacing human intelligence, but by absorbing the work of coordination, which is increasingly revealed for what it has always been: not very challenging and dull as a dishwater.

AI systems can summarize discussions, track tasks across teams, generate first drafts of reports and code, and expose anomalies and bottlenecks. And in every case, faster and likely better than a horde of junior executives. Fewer meetings. More work done.

AI will not eliminate humans or the organizations they create. It will compress hierarchies and vastly improve the corporate signal-to-noise ratio.

National productivity statistics yield averages across millions of firms, most of which have barely begun to reorganize around AI. But inside AI-native companies, the effects are already visible.

Technological revolutions rarely announce themselves through aggregate data. They begin in pockets where constraints have been removed without fanfare.

In the industrial era, the key constraint was physical production. In the computing era, it was calculation. Yet for all their power computers, left undisturbed is the constraint of coordination. (Email may have made it worse!)

The next wave of winning companies will look different from anything we have seen in the past. They will be smaller than expected and faster than expected. And I bet far more productive, per employee, than anyone has ever expected.

The AI productivity boom is not missing; it is not even hidden. It’s just happening where the experts don’t go.

At the Gilder Technology Report, we try to remember that in technology investing, the most important signals are rarely the ones making the headlines.