The Fallacy of Productivity Gains (and Why Agentic AI Won’t Save Us)

On the one hand, bad management pushes for raw speed without investing in processes. In turn, this yields a negative correlation: more output, worse outcomes, and burned-out staff. Good management, on the other hand, redesigns work to prevent errors and waste. That approach ultimately yields a positive correlation. In other words, agentic AI is not a substitute for good management. It is a mirror that is increasingly held up to bad management. Deploy it into a broken process, and you simply automate broken outcomes at machine speed.

The Ethical Fault Line

This, then, is the ethical fault line. When a vendor promises a 30% productivity gain from agentic workflows, what they are really promising is that you can fire the human who currently absorbs process friction. Further, they are promising that a large language model-based agent, with its well-documented propensity for hallucination and brittle reasoning, will somehow navigate the messy, tacit, context-dependent reality of knowledge work better than a trained professional. They are promising that you can have it fast, cheap, and good—all three, the magic triumvirate. You cannot, you never could, and you never will.

The only ethical path forward for agentic AI in KM is to invert the question. We must stop asking, “How much faster can we make this worker?” Instead, start asking, “How much better can we make this process before we add (or decide not to add) autonomy?” This means using the agents not to replace judgment, but to handle the deterministic, low-value toil that currently distracts from judgment. It also means we measure success not by keystrokes saved, but by errors avoided and cognitive load reduced.

If we have learned anything from the productivity fallacies of the past 30 years, it is that technology alone has never delivered, and likely never will, the promised quantum leap. It is always the marriage of process, governance, and human skill. Vendors will undoubtedly keep making their incredible claims because they know buyers are desperate, and frankly, it’s a marketing tactic that has made many of them rich to date. But as practitioners, we have a responsibility to remember, or learn afresh for that matter, the lessons of the late 1990s: Speed without quality is just a faster path to failure. And no agentic system, however clever, can automate its way out of that.