Three separate reports this month paint a bleak picture of generative AI inside companies: a survey of 3,750 workers across 14 countries found 54% bypass corporate AI tools and do the work manually; more than a third of U.S. employees say they received AI-generated slop in the last month; and a fresh breakdown of MIT data confirms 95% of AI initiatives fail to produce a positive return. Yet the failure, a growing chorus of experts argues, is less about the technology and more about a decades-long hollowing out of management competence.
AI slop is a symptom, not the root cause
Forty percent of U.S. workers surveyed said AI slop made up 15% of the content they encounter. When slop runs sequentially through a business process, errors compound, trust evaporates, and early productivity gains vanish. The raw hallucination rate is high enough that it seems easy to conclude generative AI simply is not ready for serious work.
Jeffrey Funk, a technology analyst, notes that the frequency of hallucinations does not block AI use everywhere - it just pushes adoption toward domains such as coding, where hallucinations can be caught and corrected. The more urgent question is why so many companies cannot manage that distinction. The answer, he suggests, lies far upstream of the model itself.
The top-down AI push is alienating workers
Multiple surveys show a sharp optimism gap: CEOs and board members are far more bullish on AI than the employees who are expected to use it. Instead of pulling AI into the workplace, as workers did with the web in the 1990s, frontline staff today are resisting tools they did not request. Google employees flooded internal boards with anti-AI memes; Microsoft canceled most of its direct Claude Code licenses; Amazon teams inflated AI token consumption to hit targets.
Cory Doctorow, in his new book "The Reverse Centaur's Guide to Life After AI," described the reversal: "If you look back to the business press of the 1980s and the late '90s, it's full of hand-wringing editorials about how bosses will cope with workers who are smuggling in the web." Today, the same outlets are filled with advice on punishing workers who refuse to use AI. This shift from pull to shove is not a technology problem, it is a leadership one. The disconnect is a focus of AI for Executives & Strategy training, which helps leaders align expectations with what actually happens on the ground.
Centaurs and reverse centaurs: who controls the tool
Doctorow's book introduces a pair of science-fiction concepts that explain the split in AI success. "Centaurs," he said, "are workers who are assisted by technology and who decide how that technology is going to assist them." These are the people who report that AI makes their lives better because they chose the application and they know the domain.
On the other side are what Doctorow calls reverse centaurs: "workers who are being asked to produce more with AI at the expense of quality, at a higher speed, at the expense of their own wellbeing, and who understand that they're being recruited to take the blame when the AI screws up their job." For reverse centaurs, the AI is in control; the human is just a monitor. The difference in outcome, Doctorow told ArsTechnica, is not determined by the gadget but "who it does it for and who it does it to."
Consultants and the hollowing out of internal expertise
The impulse to impose AI from the top often comes with heavy reliance on external consultants. Mariana Mazzucato and Rosie Collington's book The Big Con documented how decades of outsourcing have left companies and governments with fewer internal experts who can vet new technology. With AI, the problem deepens: traditional technical gatekeepers have been bypassed as AI labs court CEOs directly.
The consulting industry's own travails underscore the weakness. Accenture's stock has fallen 51% this year, and peers such as IBM and Infosys have been hit as investors question whether big consulting firms have the operational experience to ask AI the right questions or verify answers. "Real AI implementation requires deep domain expertise," critics said, a resource that has been steadily stripped from the organizations paying for it.
Why this matters for management
The mounting failures are not a verdict on generative AI itself. They are a verdict on a management style that treats AI as a mandate to be enforced rather than a capability to be cultivated. The few success stories come from teams where domain experts choose how the AI assists them - centaurs, not reverse centaurs.
For managers, rebuilding internal competence is the prerequisite. Without deep operational knowledge inside the firm, no AI pilot will be judged correctly and no output will be trusted. The shift requires investing in the people who do the work, not in dashboards that spy on them. Resources on AI for Management offer practical approaches for leading AI adoption that gives control to the people who will be accountable for the results.
Your membership also unlocks:
