Greg Brockman's five-to-ten-year bet on the end of routine computer work

Greg Brockman has put a time horizon on one of the bigger bets in technology right now. In his telling, it is not unreasonable to expect that within five to ten years, no one will be performing the routine computer tasks that currently define most knowledge work. “It’s not crazy to think that in 5 years, 10 years, no one will be doing any of this stuff anymore. That we will get our time back.” That is the call. The clock starts now.

The prediction concerns something that cuts across almost every sector: the hours spent clicking through menus, entering data into spreadsheets, and learning software that exists only to mediate between a person and a goal they actually care about. The implicit claim is that the current relationship between humans and computers is a phase, not a permanent condition. The current era, far from being the endpoint of how people interact with machines, is an artifact that AI will eventually make obsolete.

The stakes of the call being right are significant on several dimensions. For workers whose roles center on mediation tasks of this kind, the prediction implies displacement that is harder to absorb than previous waves of automation. Clicking and typing through enterprise software is not a niche skill. It is the dominant mode of white-collar labor. If that mode disappears within a decade, the retraining and structural adjustment required would be substantial, and the timeline Brockman gives is short.

It's not crazy to think that in 5 years, 10 years, no one will be doing any of this stuff anymore. That we will get our time back. Greg Brockman

For the software industry, the prediction is equally pointed. Products whose value proposition is largely the interface itself, the menus, the workflows, the dashboards, would face a direct challenge. If the interface is the thing that disappears, then any business built around optimizing or selling that interface is exposed. Brockman does not name companies or products in the claim, but the implication runs straight at the software layer that sits between enterprise data and enterprise action.

Several things would have to be true for the prediction to land on the short end of its range. AI agents would need to reach a level of reliability that organizations trust with consequential tasks, not just drafting or summarizing. The infrastructure for persistent agents that maintain context and act autonomously over time would need to mature. And adoption curves in enterprises, which have historically lagged consumer behavior by years, would need to accelerate substantially.

The prediction is falsifiable, which is what makes it worth tracking rather than dismissing. By 2031, at the five-year mark, the share of knowledge workers who still spend the majority of their day inside traditional software interfaces will either have dropped sharply or it will not. Brockman is not hedging the direction, only the timing. The range he gives, five to ten years, is wide enough to cover significant variation in adoption speed, but narrow enough that anyone working in software product, enterprise information technology, or workforce development should treat it as a planning constraint rather than a distant hypothetical. Whether the prediction proves right, early, or simply directionally correct but slow, the question it raises about what computer interaction actually needs to look like is already changing what gets built.