Employees Are Getting Faster, So Where Is the Company’s Money?

Imagine a factory in the late 19th century, billowing dark smoke just as electricity was first being introduced. At the center of the factory sat a massive steam engine, and the power it generated was distributed to individual machines through a complex web of ceiling-mounted shafts and belts.

When electric motors first arrived, the innovation strategy of most companies was remarkably simple. They removed the old steam engine from the center of the factory and replaced it with a shiny, massive electric motor. The source of power changed, but the ceiling shafts, the belts, the layout of the machines, and the workflow of the laborers remained exactly the same. Naturally, the magical productivity gains they expected never materialized.

The true potential of electricity only exploded when companies adopted the "unit drive" system—attaching small, individual electric motors to each machine. No longer forced to arrange machines around a central rotating shaft, companies could completely tear down and redesign their factory layouts to optimize the actual flow of production. The absurd inefficiency of halting the entire factory just to fix a single broken machine disappeared. Economic historian Paul David cited this as a prime example of how the value of a General Purpose Technology (GPT) is only realized when combined with a fundamental redesign of the existing production system [David, 1990].

More than a century later, companies around the world face the exact same question in the wake of the massive wave of Artificial Intelligence: Are we simply swapping out the old steam engine for a state-of-the-art AI motor, or are we redesigning the way our company works for the AI era?

Individual Productivity Gains Are Undeniable

To be clear, at the individual "task" level, AI’s ability to boost productivity is no longer up for debate. A research team led by Stanford Professor Erik Brynjolfsson observed 5,172 customer support agents at a software company. When equipped with a generative AI tool, the number of customer inquiries resolved per hour increased by an average of 15%. What’s more fascinating is that novice or low-performing employees benefited the most, seeing their productivity jump by nearly 30% [Brynjolfsson, Li & Raymond, 2025].

An experiment involving 758 management consultants reached the exact same conclusion. When assigned tasks within AI’s capability range, consultants using AI finished their work 25% faster, and the quality of their output was significantly higher. (Of course, for tasks outside AI’s capabilities, the accuracy rate actually dropped, highlighting a clear limitation) [Dell'Acqua et al., 2023]. The evidence is overflowing: when used in the right places, AI dramatically elevates human efficiency.

So, let’s ask a very logical question: "If our employees' productivity has increased by 30%, shouldn't our company's profits also increase by 30%?" The confusion we are experiencing right now starts exactly with this question.

Faster Employees vs. Stagnant Corporate Vaults

The customer service study mentioned above measured "resolutions per hour." It is a fantastic achievement, but the researchers themselves drew a line, noting that this metric does not necessarily prove a reduction in the company's total labor costs or a structural increase in profits. There is a surprisingly massive crevasse between a task getting faster and a company actually making more money.

Suppose Manager Kim, who used to take two hours to write a report, now finishes it in one hour thanks to AI. His task productivity has increased by exactly 100% (2x). But the company doesn't cut Manager Kim's salary in half. The labor cost remains the same. If Kim uses that saved hour to prospect new clients or brainstorm core strategies, the company's revenue will likely grow. But what is the reality? He might get pulled into another trivial meeting or waste that saved hour fact-checking the plausible-sounding hallucinations generated by the AI. Ultimately, the metric of "report writing time cut by 50%" is achieved, but the numbers on the financial statements don't change by a single cent.

This is precisely why we must strictly distinguish between "AI Productivity" (the speed at which work is done) and "AI Payoff" (when that speed translates into actual economic results for the company).

This gap is blatantly obvious in macro-level corporate research. According to a review of firm-level AI research up to 2026 by Professor Tania Babina, while AI's effects are prominent in narrow, well-defined tasks, those effects become like a mirage when scaled up to the enterprise level. This is because new technology is deeply intertwined with organizational structure, complementary investments, and adoption methods [Babina, 2026].

A study analyzing AI adoption in U.S. manufacturing even showed a "J-curve" effect, where companies adopting AI initially saw their productivity and profitability worsen before eventually rebounding much later [McElheran et al., 2025]. The "adjustment costs" of forcing new technology into existing processes are substantial. While recent corporate data finally shows AI investments leading to actual performance, researchers agree that the secret isn't simple headcount reduction. It lies in the accumulation of "Organization Capital"—building entirely new ways of working tailored to AI.

The Five Steps to Actual Cash Flow

Companies that adopt AI usually look at the first step, Usage (number of licensed users, prompt inputs), and feel proud. Then they measure the second step, Task Performance (how fast and how well the work was done), and declare victory.

But executives must cast their gaze higher. The third step is Process Performance. Did writing the report faster actually shorten the company’s overall decision-making cycle? The fourth step is Economic Performance. Did that accelerated process truly reduce costs, expand capacity, or generate new demand? And finally, all of this must lead to the fifth step: an increase in Cash Flow and corporate valuation.

Usage → Task Performance → Process Performance → Economic Performance → Cash Flow

If even one of these five steps is broken, an optical illusion occurs. Employees might be busy using AI and working faster, but the bottleneck in the overall process remains unchanged. If you have extra capacity but no new demand in the market, revenue stays flat. Worse, you might just end up adding expensive AI license fees and the cost of the security department required to manage it.

The Ultimate Question Executives Must Ask

Until now, we have looked at AI and asked, "How well are our employees using it?" and "How much time did we save?" But the fierce question executives truly need to ask is this: "Through what exact path did that change in productivity turn into cash in our company's vault?"

The moment you relentlessly ask this question, the vague reality of AI investments reveals its skeleton. Instead of relying on empty numbers like "20% productivity increase," you begin to strategize about which process bottleneck has been cleared, and where to redirect the freed-up resources to drive revenue and mitigate risk.

A century ago, placing a giant electric motor in an outdated factory didn't spark a productivity revolution. The magic only began when the entire workflow was rewired for electricity. The AI era is exactly the same. Handing out ChatGPT accounts to employees is merely swapping out the motor. True innovation requires redesigning your company's workflows, resource allocation, and performance measurement systems down to the bone to harness the power of AI.

To achieve this, companies must continuously make four critical decisions: Where to find value (Map), Where to focus resources (Allocate), How to prove the results with data (Prove), and What to grow or kill based on evidence (Scale/Stop).

In the upcoming installments of this series, we will dissect the framework of these four decisions one by one. In our next episode, we will peer into the most treacherous crevasse on the value path—the gap between Task Performance and Process Performance. If AI really saved our employees' precious time, where on earth did it all evaporate to?

Babina, T. (2026). Understanding firms' AI efforts and their economic impact (NBER Working Paper No. 35123). National Bureau of Economic Research. https://doi.org/10.3386/w35123

Babina, T., He, A. X., & Jiang, R. (2026). Canaries in the gold mine: Early productivity gains from artificial intelligence creating organization capital (NBER Working Paper No. 35684). National Bureau of Economic Research. https://doi.org/10.3386/w35684

Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044

David, P. A. (1990). The dynamo and the computer: An historical perspective on the modern productivity paradox. The American Economic Review, 80(2), 355–361.

Dell'Acqua, F., McFowland, E., III, Mollick, E., Lifshitz, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science. Advance online publication. https://doi.org/10.1287/orsc.2025.21838

McElheran, K., Yang, M.-J., Kroff, Z., & Brynjolfsson, E. (2025). The rise of industrial AI in America: Microfoundations of the productivity J-curve(s) (CES Working Paper No. 25-27). U.S. Census Bureau, Center for Economic Studies. https://www2.census.gov/library/working-papers/2025/adrm/ces/CES-WP-25-27.pdf