Agents or Humans? Well, Humans Are Cheaper
May 11, 2026
Companies want to divest themselves of human workers so they can save money and invest in AI.? ? Unfortunately AI is still years away from totally replace humans and Jeff Raikes of Fortune explains how in the article, “AI Is Capturing Cognition-And Most Companies Are Building A Talent Debt They Don’t See Yet.” Raikes worked at Microsoft for twenty-seven years and he never thought he’d see technology that would replicate human cognition.? ?
He explains that companies losing opportunities for their bottom line and helping people advance:
“And companies are moving too fast to fully exploit its potential without considering what they’ll lose when humans are taking a backseat to technology in the workplace…A society in which fewer people develop the capacity for independent, critical thought is not just less competitive. It is more vulnerable to manipulation, to misinformation, and to the erosion of the informed citizenship that democracy depends on.”
Entry level jobs are changing and/or being phased out because of AI. Experienced workers are seeing steady job growth. Raikes advises companies to slow down and reclaim the cognitive apprenticeship model. In other words, people need to be AI literate and use it to take their knowledge further rather than replacing humans. He suggested implementing AI literacy in schools like other countries are doing.
There is a slightly different approach to consider. “Why Humans Are Still More Cost-Effective Than AI Compute” states:
We’ve reached a point where the cost of compute can exceed the cost of human labor. AI infrastructure requires significant power and cooling, driving up operating costs. The Tech News World essay asserts:
When a human makes a mistake, it is typically limited in scope. They might miss a typo or miscalculate a cell in a spreadsheet. When AI systems fail, they can do so at scale, producing large volumes of incorrect or harmful output quickly. The “human discount” exists because we have a built-in sense of common sense that prevents us from making clearly flawed decisions. AI currently lacks this “sanity tax,” making the potential liability costs of AI much higher than the salary of a supervised human. The cost of managing AI errors and liability is still not fully reflected in most analyses. We haven’t yet priced in the cost of hallucination insurance.
Here’s a test. Just ask a person how the AI system is working at their firm.
Whitney Grace, May 12, 2026
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
