Reports of AI corporate implementation failures have continued to mount over the last year. Published last fall, an MIT Study found that 95% of AI initiatives fail to achieve a positive ROI. A subsequent study from software company Atlassian reported similar levels of failure.

This year, even tech company employees are rebelling against AI or at least not cooperating. Google “employees were sharing heaps of anti-AI memes” on internal messaging boards. A Nvidia exec said “AI is more expensive than actual workers.” Amazon employees “were caught inflating AI token consumption to hit internal usage targets” and Microsoft began “canceling most of its direct Claude Code licenses. And a new global survey of 3,750 executives and employees across 14 countries “finds that more than 54% of workers bypassed their company’s AI tools in the past 30 days and completed the work manually instead.”

AI Slop is One Problem

One reason for the failures and worker resistance is AI slop. Forty percent of survey respondents (1,150 U.S.-based full-time employees) said they had received slop in the previous month and that 15% of the content they receive was workslop. A more recent analysis says that when slop occurs in sequence across a business’s processes, those processes and their outputs start to deteriorate, errors compound, trust erodes, and any productivity gains disappear.

These studies make it easy to conclude that generative AI is terrible, and it must be stopped. The frequency of AI slop is too high for AI to be successful and there are few solutions, except to wait until the AI labs produce better AI products.

There is another explanation, however. Although AI slop is a major problem, the frequency of hallucinations doesn’t prevent generative AI’s usage, it merely pushes us towards those applications (e.g., coding) where the hallucinations can be better managed and towards a better process of implementing and managing AI.

Why isn’t AI being managed better? One problem is that CEOs and other top managers, influenced by extraordinary claims from tech leaders, big consulting companies, and elite professors, have tried to ram AI down the throats of workers. Numerous surveys have found that workers are far less optimistic about AI than are top managers, and board members are the most optimistic. In other words, the further you are from real work, the more optimistic you are about AI.

Cory Doctorow: “The Reverse Centaur’s Guide to Life After AI.”

Cory Doctorow sheds light on AI’s failure in his new book, which follows his best-selling “Enshittification,” and in an interview with  ArsTechnica. Doctorow reminds us that “workers actually wanted earlier technological breakthroughs and often had to fight to get them into the workplace. With AI, people are more likely to feel that the technology is being shoved down our throats; some workers are even required to use it.” He says:

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. You look at those same press outlets today, and it’s full of people saying, ‘What are we going to do about the fact that no one in the workplace wants to use AI?’—along with ads for firms that will spy on your workers for you so that you can punish the workers who refuse to use AI.

While this makes generative AI sound as though it could never work, He also describes some thoughtful, sensible defenders:

Sone of the paradoxes that I try to explore in this book is the workers who are not fools, who are historically good, reliable narrators of their own experience, and who tell you that AI is making their lives better because they have the capabilities to do so and have found the right application. One interpretation is that AI can work if it is done by the right people.

He describes the paradox in terms of science fiction in which “what the gadget does is less important than who it does it for and who it does it to.” He calls those people “centaurs,” a human with a horse’s body:

“They are workers who are assisted by technology and who decide how that technology is going to assist them. Whereas the workers who hate it are 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. They are called reverse centaurs.

The centaurs understand that statistical inference using convoluted deep neural networks is not bad while reverse centaurs don’t think, they are monitored and controlled by AI.

Corporate leaders shoving AI down the throats of workers reflects a return to top-down management this time using external consultants, whose numbers have exploded over the last 50 years. As Mariana Mazzucato and Rosie Collington say in their 2023 book, The Big Con, “the more governments and businesses outsource, the less they know how to do.”

Put simply, companies have reduced the number of internal experts over the last 20 years and with AI they are listening to internal people even less. Combined with the strategy of Sam Altman and other tech leaders to sell AI to CEOs and other elites, traditional technical gatekeepers have been bypassed.

Even the stock market has recently concluded that big consulting companies don’t know enough to effectively implement AI. Accenture’s stock has fallen 51% this year and other companies that sell technological services have been affected, too, including IBM and Infosys. “Accenture’s growth hinges on helping companies implement AI, but investors are increasingly unsure if the firm can come through on these commitments.” Why? Because many critics say that “consultants at Accenture and other large firms generally lack the operational experience required to ask AI the right questions or verify if its outputs are correct. Real AI implementation requires deep domain expertise.”

Changes in the Startup System

Running parallel to the increasing reliance on external consultants are changes in America’s startup system, which have caused AI labs to emphasize hype and moonshots. Forty years ago, startups found a niche, became profitable, and did an IPO. As they slowly expanded their market, their market capitalization increased.

For the last 20 years though, there has been a move towards making extraordinary claims, getting big valuations, and then doing an IPO usually without profits. SpaceX, OpenAI, and Anthropic are just the latest in a series of over-valued Unicorns that I documented in my 2024 book, Unicorns, Hype and Bubbles. There is still only one startup (Uber) founded since 2005 that is both profitable and among the top 100 companies in terms of market capitalization, compared to 23 founded between 1975 and 2005.

The upshot is that AI slop might not be the biggest problem faced by America right now; it might be the decline in America’s startup system and the hollowing out of corporate competencies. If this is true, the capitalists need to save capitalism before others (e.g., Zohran Mamdani) impose their solutions.