The Paradox of Automation: Why AI Increases Human Work
the gist
Dan Shipper argues that AI doesn't replace human work; instead, it commoditizes expert competence, creating a glut of 'near-correct' output that requires more human experts to refine, direct, and curate.
The Automation Paradox
Dan Shipper, founder of Every, argues that the common fear of AI-driven mass unemployment is based on a misunderstanding of how AI functions in the workplace. While AI can perform tasks that previously required expert-level competence, it does so by synthesizing existing data, resulting in output that is often 'close but not quite right.' This creates a paradox: as automation makes basic expert tasks cheaper and more accessible, the demand for human experts actually increases. These experts are needed to curate, refine, and shepherd AI-generated work into final, usable products.
The Role of the Expert
In an 'agent-native' company like Every, AI agents are ubiquitous, yet the company has grown significantly in headcount. Shipper explains that AI acts as a force multiplier for non-experts, allowing them to attempt tasks they previously couldn't. However, this floods the organization with 'near-correct' work. The critical human role shifts from performing the initial task to acting as an editor, system architect, and quality controller. Experts are now responsible for building the 'rails'—the review processes and guidelines—that ensure AI output meets professional standards.
Agency vs. Autonomy
Shipper distinguishes between an agent's ability to act (autonomy) and true human agency. While agents can be programmed to perform complex, multi-step tasks, they lack the self-motivated, playful, and value-driven nature of humans. He argues that even as agents become more capable, they remain fundamentally dependent on human direction. They are tools that 'look back' at the user to ask, 'What should I do next?' This dependency ensures that humans remain the ultimate decision-makers, regardless of how advanced the underlying models become.
The Future of Work
Shipper remains optimistic about AGI, defining it as an agent so valuable that it makes economic sense to keep it running continuously. He contends that the fear of AI layoffs is often a reaction to the initial shock of seeing a model perform a high-level task, rather than a realistic assessment of long-term integration. He suggests that the rate of organizational change is slower than people expect, and that the most successful individuals will be those who 'ride the models'—using them to expand their own capabilities rather than viewing them as a replacement for their own judgment.