AI capability is increasingly augmenting, and in some cases replacing, portions of human work across the enterprise. That possibility was anticipated almost immediately after AI entered mainstream business conversations. What remained uncertain was the speed and scale of the impact, both of which appear closely tied to organizational adoption rates, implementation maturity, and the degree to which AI-enabled processes can be trusted to produce reliable, repeatable, and sustainable outcomes.
The situation feels reminiscent of the introduction of the moving assembly line in manufacturing. The comparison isn't perfect though. Assembly lines transformed physical production while AI is transforming knowledge work. Still, the underlying economics seem surprisingly similar. In both cases, organizations are applying technology to improve productivity, consistency, quality, throughput, and cost efficiency. Both introduced concerns about worker displacement. Both fundamentally changed the relationship between labor, production capacity, and business investment.
The corporate workplace has always functioned as a knowledge factory. Organizations hire talented people to apply expertise, judgment, and effort toward outcomes the business values. As AI systems become capable of assuming portions of that work, it seems natural that many organizations would evaluate AI investments through a labor lens.
FTE Math Seems Right
The labor math is easy to understand and it seems to track. If an organization spends a certain amount on AI technology and subsequently avoids hiring ten people, the natural comparison then is between the cost of the technology and the cost of those ten people. If the technology costs less than the workforce capacity it displaces, the investment appears justified.
The problem is that AI capability doesn't necessarily behave like labor capacity. Human capacity scales primarily through time. Adding another employee generally adds productive hours available to perform work. Productivity may not scale perfectly due to coordination and management overhead, but labor is still typically measured in hours, headcount, and utilization.
Digital capability operates differently. AI systems are constrained by architecture, utilization, throughput, workflow design, service quality, and demand rather than available labor hours. Their value isn't necessarily reflected in the number of people they replace. It may be reflected in the amount of work they process, the quality and consistency of the outcomes they produce, the rate of rework they generate, and the cost required to deliver those outcomes.
But AI Doesn’t Behave Like Labor
Viewed through that lens, AI capacity starts to look less like labor and more like a production asset. To understand its value, we would look to measure the cost of the capability, the volume of work it can perform, the quality of its outputs, the frequency of errors and rework, and the business outcomes it contributes to. Combined, those measures allow us to understand production rates, service costs, and ultimately cost of delivering an outcome.
Instead of asking how many people the technology replaces, we begin asking how many units of work it performs. What is the cost per outcome? How much rework does it eliminate or create? What level of utilization is required to achieve economies of scale? How does increasing capacity affect service quality, customer outcomes, or operational performance?
Which Raises an Interesting Measurement Question
The uncomfortable dimensions of this transition are worker displacement and the broader macroeconomic effects of transferring portions of productive capacity from people to technology. There is, however, another implication that receives far less attention.
The accounting treatment of digital productive assets.
In traditional manufacturing environments, machinery is typically capitalized because it contributes productive capacity over multiple periods. The asset itself is recorded on the balance sheet and depreciated over its useful life. As production occurs, a portion of that depreciation becomes part of the cost structure associated with producing goods and ultimately contributes to cost of goods sold when those goods are sold.
AI doesn't fit neatly into that model. We don't typically think about work in terms of units of production or services consumed within the enterprise. We think about people, departments, and budgets. Measuring digital capability as productive capacity requires a different perspective.
Many enterprise AI investments are consumed through subscriptions, cloud services, and usage-based pricing models that are recognized primarily as operating expenses. Yet economically, many of these capabilities behave less like traditional service consumption and more like production infrastructure. They perform work, produce outcomes, and create capacity that organizations come to depend upon in much the same way they depend upon manufacturing equipment.
What if We’re Measuring the Wrong Thing?
If AI increasingly functions as productive capacity rather than labor capacity, are we measuring it using the wrong economic framework?
The production-capacity model requires organizations to understand something very different from headcount. We would need to understand the cost of the capability, how utilization can be measured, how units of work can be defined, how outcomes can be associated with capacity consumption, and how costs can be allocated across products, services, and business capabilities in ways that are repeatable and auditable.
That approach is more involved than applying traditional FTE-based calculations to digital capability, but, it may also be considerably more accurate. The long-term question may not be how many employees an AI system can replace. It may be whether labor is the right unit of measure at all.
If AI ultimately represents a form of digital productive capacity, then applying workforce economics to it may be no more effective than managing a manufacturing plant by counting the number of people standing next to the machines. At some point we need to understand the machine itself: what it costs, how much work it performs, how consistently it performs that work, and how the value it creates is reflected in the economics of the business.
