Ideas
The Future of Work Is a Choice
AI can double productivity with the same workforce, or deliver the same output with half the people. Which outcome we get is not determined by the technology.
Last updated: · 3 min read
The displacement narrative treats the future of work as something that happens to people. The augmentation narrative treats it as something people can shape. Both are partly right. The outcome depends on choices by policymakers, technologists, employers, and workers themselves. The question is who gets to make those choices.
Too often, the conversation about AI and work gets framed as a prediction: how many jobs will AI eliminate? When will automation replace human workers? Those questions assume the outcome is determined by the technology. It isn't. The outcome is determined by the choices we make about how to deploy it.
Here's the fork in the road: do we want AI to double productivity with the same number of people, or deliver the same output with half the workforce? Both are technically possible. One path treats workers as a cost to be reduced. The other treats human capability as something to be strengthened. One concentrates the gains of AI among the people who own the systems. The other distributes those gains across the people who use them. The technology doesn't choose between these futures. We do. And right now, the default - markets optimizing for short-term returns - is choosing the first path by inertia.
The care economy is where this choice becomes most visible. There are 4.3 million home health and personal care aides in the United States. The number of Americans over 65 will grow from 58 million to 82 million by 2050, and the demand for human care workers will grow with it. These are jobs that AI cannot replace - the physical presence, the emotional judgment, the trust that develops between a caregiver and the person in their care. But AI can make those jobs better. Real-time assistance for workers managing complex medical equipment. Monitoring tools that flag risk before a crisis develops. Scheduling systems that reduce the administrative burden that drives burnout. The question is whether we design AI to augment these workers or to surveil and manage them. That's a design choice, and it will be made by whoever controls the development process.
I co-authored a piece in US News with Jade Lin making the case that AI policy for work has to start from the workers, not the technology. Policymakers and technologists cannot be the only ones who decide how AI is used and what values are embedded into it. The tools being built for home care should be developed in consultation with care workers. The same principle applies across industries - in agriculture, in education, in logistics. The people whose working lives will be reshaped by AI should have authority over how that reshaping happens. That's a governance argument, not a labor argument, and it connects directly to everything else I work on.
The WEF Future of Jobs Report 2025 found that 86 percent of employers expect AI to transform their business by 2030, and 41 percent plan to reduce staff in roles where skills are becoming less relevant. At the same time, care economy jobs - nursing, social work, personal care - are among the fastest-growing categories globally. That divergence tells you something about where the real economic and social value lies in an AI-enabled economy. The jobs that are hardest to automate are often the ones our economy values least. AI gives us an opportunity to correct that, if we treat the question as one of institutional design rather than market optimization.
At the McGovern Foundation, we fund organizations working at this intersection. Digital Green's AI-powered advisory tools don't replace agricultural extension workers - they make those workers more effective, reaching hundreds of thousands of farmers who previously had no access to expert guidance. Khushi Baby doesn't replace community health workers in Rajasthan - it gives them data infrastructure that makes their knowledge actionable at scale. The pattern is the same across our portfolio: AI that treats human workers as the point of the system rather than the bottleneck in it.
The phrase I keep coming back to is simple: don't adopt AI to do things better - use AI to do better things. The distinction sounds small. In practice, it's the difference between an economy that uses AI to squeeze more out of fewer people and one that uses AI to extend human capability into places it couldn't previously reach. Left to markets focused on short-term profit, displacement will dominate. Getting the other outcome requires policymakers, technologists, and civil society to design for it deliberately. The future of work isn't a forecast. It's a decision we're making right now, whether we realize it or not.
Related essays