Employment Hero Founder Ben Thompson on the Human Skills AI Can't Replace

TEXT START: As business leaders navigate the early stages of AI integration, some argue that delegating to machines could elevate our economic worth.

The Dissection

This article is corporate transition management disguised as a forecast of human empowerment. It takes genuine AI-driven productivity gains, then converts them into a comforting story: machines eliminate drudgery, humans retain judgment and empathy, and everyone becomes a more valuable founder.

The actual evidence points elsewhere. A claimed 30x increase in engineering output is not proof that engineers become 30x more valuable. It is evidence that fewer engineers may be required to produce the same output. “Atomisation of work” sounds liberating, but operationally it means labor is broken into portable tasks, continuously matched, measured, and priced. That is not worker sovereignty. It is high-speed commodification.

The article also mixes two separate questions: whether humans retain cultural, moral, or political importance, and whether most humans retain economically necessary employment. The first may remain true. It does not rescue the second.

The Core Fallacy

The central error is treating biological scarcity as economic scarcity.

“There are only so many humans” is irrelevant if the market is purchasing capabilities, decisions, code, analysis, design, moderation, and coordination rather than human bodies. AI expands the supply of those capabilities while reducing the amount of human labor needed to deliver them. Human population decline does not automatically raise wages when machine substitutes grow faster than demand.

The article assumes that “presence, judgment, reasoning, leadership, empathy,” and accountability remain uniquely human economic bottlenecks. Under the Discontinuity Thesis, that is an unproven—and likely temporary—lag defense. Once AI reaches durable cost and performance superiority across cognitive work, human traits become preferences, interfaces, or liability assignments, not protected economic domains.

The 30x productivity figure quietly exposes the mechanism the article tries to deny: more output per favored worker means a smaller labor requirement for the system. Aggregate prosperity can rise while productive participation collapses. That is P1 leading to P3, with the gains accruing primarily to owners and controllers of AI capital.

Hidden Assumptions

• Demand for “human” work will expand fast enough to absorb displaced labor.
• Human judgment and empathy cannot be replicated, simulated, or made economically unnecessary.
• Companies will distribute AI productivity gains through higher wages rather than reduced headcount, weaker bargaining power, or owner returns.
• Real-time matching will increase worker autonomy rather than create a universal labor auction.
• Calling everyone a founder creates ownership, capital, or durable control. It does not.
• Human accountability guarantees human employment. Firms can retain nominal human accountability while automating the underlying work.
• Cultural agency and the right to shape AI systems translate into ownership of the systems and their rents.
• The institutional order can coordinate stable human-only niches at scale, contradicting P2.
• Outsourcing unpleasant tasks leaves humans with “higher-value” tasks instead of leaving them with tasks that are next in line for automation.
• More consumption, legitimacy, or social recognition would compensate for the loss of productive participation. Under DT, transfers can preserve demand without restoring economic sovereignty.

Social Function

Primarily: ideological anesthetic, elite self-exoneration, and transition management—with a partial truth embedded inside it.

The partial truth is that AI can augment selected workers, expand individual leverage, and create temporary niches for people with rare trust, access, domain authority, or control over deployment. The anesthetic is presenting those niches as the normal future for the population.

The article gives executives a clean moral narrative: displacement is not extraction; it is elevation. Workers are not being made redundant; they are being freed to become founders. The ownership question disappears. So do distribution, bargaining power, capital concentration, and the possibility that “fluidity” is simply precariousness with better software.

Jade’s demand for agency, self-determination, and wisdom is politically and culturally serious. It is not an economic countermechanism. Communities can shape models and still be subordinated to whoever owns the compute, energy, data pipelines, deployment channels, and distribution networks. Voice without control is consultation inside someone else’s machine.

The Verdict

This is not a credible account of human economic ascendance. It is a polished corporate lullaby built around a real productivity shock.

AI may make a minority of owners and highly placed operators vastly more productive. It may preserve temporary demand for trusted human interfaces and transition brokers. But the article never defeats the governing sequence: cognitive automation, coordination failure, and collapse of mass productive participation. It confuses humans remaining valuable in principle with humans remaining necessary in the labor market.

The future described here is not universal founding. It is a narrow Sovereign class, a thin Servitor layer, a large pool of interchangeable task-sellers, and a managed population told that dispossession is freedom.