**Dr. Ching Hsu, PhD (**AI Researcher, FluentLab)
**Wilson Tai (**Co-founder & COO, FundFluent)
Why the real AI question is not “how many people can we replace?”, but “which layer of the organization should AI occupy?”
AI does not only change how much work a team can do. It changes where judgment sits inside the company: what juniors still learn by doing, what seniors still need to supervise, and what machines can now execute on their own.
When companies talk about AI today, the first question is often:
“How many people can it help us save?”
This is not the wrong question. In an environment where AI capabilities are improving quickly, raising operational efficiency and lowering labor costs are valid reasons for adoption.
But this question is incomplete.
If we only look at AI through the lens of headcount reduction, we may miss the deeper organizational change taking place. AI may not only affect how many people a firm needs. It may also affect how work is divided inside the firm: who handles routine tasks, who handles exceptions, whose judgment becomes more valuable, and how knowledge travels across different layers of the organization.
This is the question explored by Ide and Talamas (2023) in Artificial Intelligence in the Knowledge Economy. Their paper gives us a useful way to understand a tension that now appears frequently in both research and business practice.
On one side, experiments suggest that generative AI often helps less experienced workers the most. It acts like an always-available teaching assistant, filling knowledge gaps and lowering the threshold for newcomers to complete complex tasks. Customer support assistance, writing support, and code suggestions are all examples of this pattern (Brynjolfsson et al., 2023; Noy & Zhang, 2023).
On the other side, companies are also beginning to treat generative AI as a technology that can substitute for some entry-level white-collar work, while amplifying the value of high-level talent.
These two observations seem to point in different directions:
Is AI helping less experienced workers catch up, or is it making stronger workers even stronger?
Ide and Talamas’s answer is that both can be true. The outcome depends on whether AI acts as a co-pilot or a co-worker, and on where AI’s capability level sits within the original human division of labor.
This article is our reading of that framework.
We first reconstruct the baseline model behind the paper: how knowledge organizations work before AI, and why firms naturally divide work between lower-level “workers” and higher-level “solvers.” We then explain how AI changes this structure when it enters different layers of the firm. Finally, we offer FluentLab’s view on what this means for business owners, startup founders, and operators.
To understand what AI changes, we therefore need to begin with the world before AI.