The Gap Between Could and Does
This week brought one of those rare moments where a single research report crystallizes what dozens of headlines have been hinting at for months. Anthropic, the company behind the Claude AI system, published a study measuring how large language models are actually affecting the labor market. Their core finding is both reassuring and unsettling: AI could automate far more work than it currently does. The gap between theoretical capability and actual workplace adoption is enormous.
That gap is the story of 2026. And it tells us something important about what happens next.
The Numbers Are Telling
Anthropic’s researchers examined nearly 800 occupations and found that computer programmers, customer service representatives, and data entry workers face the highest “observed exposure” to AI tools. For programmers, AI can already handle roughly 75 percent of task coverage. Yet actual adoption in those roles remains a fraction of what is technically possible.
Meanwhile, a Dallas Fed report found that AI is making entry-level jobs harder to land. Job finding rates for younger workers entering AI-exposed occupations have dropped about 14 percent since ChatGPT launched. The jobs are not vanishing overnight. They are quietly getting harder to break into.
And then there are the layoffs. A LiveMint tracker reports that Oracle, Amazon, and Meta have collectively cut over 35,000 jobs in 2026 alone, many explicitly linked to AI-driven restructuring. The New York Times ran an opinion piece this week titled “Mass Hysteria. Thousands of Jobs Lost. Just How Bad Is It Going to Get?” that examines how AI’s impact on white-collar work may transform both politics and society.
The pattern is clear. AI is not replacing entire workforces in a single sweep. It is restructuring them selectively, starting with the most routine knowledge tasks and slowly working its way up the complexity ladder.
The Liberal Arts Counterargument
Against this backdrop, a Washington Post opinion piece by Assumption University president Greg Weiner made a provocative argument: AI’s best use is enhancing human judgment, so study liberal arts. Weiner contends that higher education is learning exactly the wrong lesson from AI’s workplace transformation. Universities are rushing to build AI programs and technical certificates when they should be doubling down on the disciplines that teach people how to think, interpret, and judge.
This is not an anti-technology position. It is a recognition that when machines handle the execution, the value shifts to the person who decides what should be executed and why. The person who can read a situation, understand context, and make a judgment call that no algorithm would reach.
Deloitte’s 2026 Global Human Capital Trends report, also released this week, reinforces the point from a different angle. Their survey finds that AI is “increasingly blurring authorship and eroding confidence from both workers and organizations.” Few companies are making significant progress addressing this. The tools are getting better. The human frameworks for using them wisely are not keeping pace.
Where Human Value Actually Lives
Here is what all five of these stories share: they describe the same transition from different vantage points.
The Anthropic data shows us the technical capability gap. The Dallas Fed shows us the entry-level squeeze. The layoff numbers show us the corporate response. The Washington Post makes the educational argument. Deloitte captures the organizational confusion.
What none of them fully articulate is the framework for what comes next. And this is where I keep returning to the same model.
When AI removes THE GRIND, the repetitive, routine, execution-heavy work that has defined most careers, human value does not disappear. It concentrates in four areas: Interpretation, Integration, Interpersonal connection, and Imagination. The 4I model.
Interpretation is what Greg Weiner is really arguing for when he champions liberal arts. It is the ability to look at data, a situation, a market, or a conversation and understand what it actually means. AI can summarize a thousand customer reviews. A human decides which pattern matters and what to do about it.
Integration is what Deloitte is struggling with. When AI blurs authorship and organizations cannot figure out who is responsible for what, the problem is not technical. It is integrative. Someone needs to stitch together the human and machine contributions into something coherent, trustworthy, and strategically sound.
Interpersonal connection is what separates the entry-level workers who will thrive from those the Dallas Fed data describes as stuck. The technical barrier to entry is rising. But the workers who can troubleshoot, collaborate, and navigate ambiguous human situations still get hired. Digital skills matter. People skills matter more.
Imagination is what the layoff numbers obscure. When Oracle cuts thousands of jobs to fund AI data centers, the implicit assumption is that the machines will do the work. But who decides what the machines should build? Who envisions the products, services, and strategies that AI infrastructure makes possible? That is imagination, and it is the one thing you cannot automate.
The Gap Is the Opportunity
The most important insight from Anthropic’s research is not that AI can do 75 percent of a programmer’s tasks. It is that actual adoption remains far below theoretical capability. That gap exists because organizations have not figured out how to integrate these tools into real workflows, real teams, and real decision-making structures.
Closing that gap requires every element of the 4I model. It requires people who can interpret what the AI is actually good at (and what it is not). People who can integrate AI outputs into existing business processes. People who can maintain the interpersonal trust and communication that keeps teams functional during massive change. And people who can imagine entirely new ways of working that have not existed before.
The layoffs will continue. The entry-level squeeze will get worse. The technical capabilities will keep advancing. None of that is in question.
The question is whether we are preparing people for the world that is emerging or the one that is disappearing. The gap between what AI could do and what it actually does is not a technology problem. It is a human one. And it is going to be the defining challenge of the next decade.
The grind is ending. What we build in its place is up to us.
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.
Get essays like this in your inbox
New essays on AI, education, and the future of work. No spam, unsubscribe anytime.