In 1987, economist Robert Solow made a famously dry observation: "You can see the computer age everywhere but in the productivity statistics." Nearly four decades later, Apollo Chief Economist Torsten Slok is making the same point about AI. Employment data, productivity metrics, inflation numbers, profit margins for companies outside the Magnificent 7: none of it shows AI transforming the economy. Not yet. The Solow Paradox is back, and it brought friends.

This should be reassuring. It is not.

Because while the macro data stays flat, the micro stories are getting louder by the day. Microsoft's AI chief Mustafa Suleyman told audiences this week that "professional-grade" AI could automate most white-collar work within 12 to 18 months. The Guardian is running features on white-collar workers ditching their careers for trades as AI erodes their professional footing. India kicked off its AI Impact Summit 2026 today with a blunt assessment: 50% of jobs will be automated, but 50% more will be created. The real question is whether anyone is preparing for the transition.

And then there is this, from Brookings: a new study measuring workers' adaptive capacity finds that 6.1 million highly AI-exposed workers lack the resources to navigate displacement. Limited savings. Advanced age. Narrow skill sets. Scarce local opportunities. Eighty-six percent of them are women. These workers are concentrated in clerical and administrative roles, and they cluster geographically in college towns and state capitals. The irony is sharp: the places that produce knowledge workers are home to the people least equipped to survive what knowledge automation is about to do.

Meanwhile, higher education is scrambling. Computer science enrollment is reportedly plummeting as students rush toward AI-specific degrees. The instinct makes sense. If AI is eating the world, learn AI. But this is the wrong lesson, drawn from the right anxiety. Teaching students to build AI tools is valuable, but it addresses yesterday's bottleneck. The bottleneck ahead is not "who can build the machine." It is "who can do what the machine cannot."

This is where the gap between the Solow Paradox and the Suleyman Prediction tells us something useful. The macro data has not caught up because AI, for all its capability, is still mostly doing the grind. It is summarizing documents, generating first drafts, answering routine queries, automating workflows. These are real productivity gains, but they are invisible in aggregate because they replace tasks, not jobs. Not yet.

The transition happens when organizations stop asking "how do we bolt AI onto our existing processes?" and start asking "what do we actually need humans for?" That is the question my book, After the Grind, was written to answer.

The 4I Framework offers a map. When AI handles the grind, human value concentrates in four domains:

Interpretive work. AI can surface data, but someone has to decide what it means in context. The Brookings study is a perfect example. The data on adaptive capacity exists. The policy implications require human judgment about values, priorities, and tradeoffs that no model can resolve.

Integrative work. Connecting ideas across domains, synthesizing perspectives that do not naturally speak to each other. The fact that AI enrollment is surging while adaptive capacity research highlights vulnerable populations in college towns suggests a systems failure. Someone needs to connect those dots and design solutions that address both.

Interpersonal work. The Guardian's profile of white-collar workers switching to trades is, at its core, a story about human connection. These professionals are not just learning plumbing or electrical work. They are rediscovering the value of showing up, being present, solving a problem for a person standing right in front of them. AI cannot do that. It cannot look someone in the eye and say, "I understand. Let me help."

Imaginative work. The hardest one to teach, and the hardest for AI to replicate. Imagining a future that does not yet exist. Designing institutions, curricula, safety nets, and career paths for a world where the grind is gone. Suleyman says 18 months. The macro data says not yet. Someone has to imagine what "after" looks like and start building it now.

The Solow Paradox resolved itself eventually. Computers did transform productivity; it just took longer than anyone expected, and the gains showed up in unexpected places. AI will do the same. The question is not whether the transformation is coming. It is whether we will be ready when it arrives.

For the 6.1 million workers Brookings identified, "ready" means something concrete: financial resources, retraining access, geographic mobility, and skill breadth. For the students flooding into AI degree programs, "ready" means something different: understanding that technical skill is necessary but not sufficient. The future belongs to people who can interpret, integrate, connect, and imagine.

The grind is ending. The macro data will catch up. The only question that matters is what you are building in the meantime.

Andrew Perkins is the author of After the Grind: Rethinking Your Business Career in the Age of Artificial Intelligence and Robotics and Chair of the Department of Marketing and International Business at Washington State University's Carson College of Business.

📖 Read this on the web: afterthegrind.ai/posts/2026-02-16-the-solow-paradox-returns/

📚 Get the book: After the Grind on Amazon