Where Is Your AI Productivity Going?

TL;DR

AI rollouts are working. People are visibly more productive: code, drafts, analysis, decks, all coming out faster than they did two years ago. Adoption is strong, yet for many companies the financials haven't moved.

AI productivity hasn't yet shown up consistently in company financials for several reasons (from the early stage of enterprise deployment to the experimental nature of many implementations), but this article focuses on one that receives surprisingly little attention.

Time saved by a person is not the same as value captured by a company.

When someone saves four hours, those four hours go somewhere. In almost every company, they go nowhere measurable. They become slack, more output nobody downstream asked for, or a slightly easier week. The productivity is real, but it evaporates at the boundary between the person and the firm.

So the question that matters isn't the grand one everyone poses: how do we transform the company for AI? It's narrow, and almost nobody asks it: when people save time, where does it go?

This has happened before

The last time an innovation made one step of the work dramatically cheaper, the value it promised took twenty years to arrive. The innovation was a steel container.

In April 1956, a converted oil tanker pulled out of Port Newark carrying fifty-eight steel containers and pointed toward Houston. Behind it was a trucking man named Malcolm McLean, working from an idea so plain it sounds like nothing: instead of unloading a truck's contents onto a ship and loading them again by hand, lift the whole container. Truck to ship to truck, sealed the whole way. It worked the first time out, faster and cheaper than the old way, a small army of longshoremen carrying sacks, crates, and barrels up a gangplank one piece at a time.

Yet for the next twenty years, it changed almost nothing. The container was better; the world around it was not. Ports were laid out for loose cargo, ship holds were the wrong shape for stacking, and labor contracts protected the very headcount the container made unnecessary. McLean had a faster way to move cargo and almost no way to turn that into a faster, cheaper supply chain. The savings had nowhere to go, so the system swallowed it.

The container did not remake global trade; the decade-plus of expensive, unglamorous reconstruction that came after did: purpose-built ships, cranes that moved a container in seconds, ports rebuilt from the waterline up, trucking and rail re-plumbed into a single system. The winners weren't the carriers who set a few containers on their existing ships and called it progress. They were the ones who rebuilt the port.

Marc Levinson tells the full story in The Box, but the line worth keeping is this: the container was faster from day one. The value took twenty years to arrive.

A savings at one step is not value to the whole system. It leaks unless something is built to capture it.

That's exactly where companies are now

People are the containers: more productive at their work. Companies are the 1956 port: organized in a way that productivity has nowhere to go and quietly disappears. When time gets saved, there are only four things a company can do with it:

Everything written about AI “transformation,” stripped of the jargon, is just the work of making the first three happen instead of the fourth. That's all it is. Not a forty-page strategy, but a deliberate way to convert reclaimed time into reinvested, compounded, or banked value, and a way to measure the results.

The container sat on the dock for twenty years before the ports caught up. We don't have twenty years, but the choice is the same, and sharper than it looks. The people are already on the dock, hired, equipped, and more productive than ever. The ports are already built, just the 1956 version. Companies need to rebuild around systems that capture and channel those gains, or watch them leak back out to sea.

Takeaways