Jeff Bezos, founder of Amazon, just predicted in a CNBC interview that rather than gut employment opportunities, AI is leading to an impending “labor shortage” as a result of AI. The employment doomsayers are wrong, he said. “What’s really going to happen is it’s going to elevate people.” By analogy, a software engineer who has been digging out a basement with a shovel gets handed a bulldozer. "We humans are never going to run out of problems, or the need for solutions, but the work is going to be done at a higher level, with a bulldozer."

Dan Shipper, CEO of Every, sees this within his company, and across the industry. AI speeds things up and solves problems, but the human work is nonstop. AI is "creating more work for humans, not less. The more we automate, the more expert human work there is to do.”

In his latest missive , Shipper describes how his own company has “automated every single thing we can with AI agents. And yet there’s way more human work to do than ever. We’ve gone from four to more than 30 human employees since GPT-3.”

The paradox of AI is that replacing some aspects of expert work may only accentuate the need for human experts. “It creates more situations where expert judgment is needed," Shipper explained. "When operations people submit pull requests with AI, you need engineers to review them. When marketers make YouTube thumbnails, you need designers to sharpen them. When engineers write, you need writers and editors to make the draft good.”

There are two forces at work, Shipper explains. First, AI still requires human oversight at the beginning and at the end of a process. Second, as AI churns out homogenized sameness in results, it increases the value of human expertise, to the point it becomes a form of status.

In terms of the work AI is taking on, human expertise is required for maintenance. “We have a team of AI engineers who are in charge of making sure our agents work well – and we’ll need them for the foreseeable future,” he said. Humans are needed at the start and end of every process involving agents:

  1. At the start of a process, humans “set the frame: what are we trying to do? What counts as good?”
  2. In the middle of a process, AI agents “collapses the task: drafts, searches, summarizes, compares.”
  3. At the end of a process, humans “judge and extend: Is it good? Where does it belong? What should happen next?”

Yes, there are roles in which AI agents can readily assume roles such as “code, prose, images, support tickets, product specs, and more,” Shipper says. "They take all of it—the exhaust of successfully completed tasks—and package it in a form that’s available to anyone, cheaply. The net effect is that skills that used to be rare – coding a pull request, making a YouTube thumbnail, writing a newsletter – are now broadly available to almost anyone.”

At the same time, such abundance creates sameness, which drives demand for expertise that can provide differentiation and variety. Widely available models deliver “visible sameness, repeated ad nauseam.”

That sameness creates a demand for differentiation that only humans can deliver. "When too much of it starts to look the same, we smell a rat," Shipper said.

For some aspects of business, such blandness and sameness is fine – for instance, in financial reports, or in customer surveys. But for innovation and strategic thought leadership, forward-looking businesses need output “that feels alive and specific, not cheap and generic,” Shipper says. "We want something that has status.”

That, in turn, drives greater demand for experts, Shipper pointed out. “Rare and valuable work must come from a human. The current generation of models only knows about work that has been done. Humans know about what needs to be done, right now, at this moment. Once a situation has been reduced to text, once it has become corpus, it is a corpse.”