Forbes contributors publish independent expert analyses and insights.
If you read popular business and economic media or venture into LinkedIn, you’ll notice many stories of how artificial intelligence is going to take away jobs today. Frankly, the umbrella term for a range of technologies that started coming into existence in the 1960s, already does and has been doing so for at least 10 years.
Current developments, like generative AI systems that can create text or image responses to prompts, have taken the business world by storm and have become objects of consumer fashion. However, research suggests that practicalities, like heavy project failure rates in businesses, might take some pressure off the assumption that all things are programmable.
The theory of using technology to boost business is an old one: Use machines, computers, communications, whatever is available to reduce the need for “expensive” and “unreliable” human effort. Part of the argument to reduce concern over employment has been that each new wave of technology has eliminated jobs but created more that were new, making up for what was lost. The use of tech was supposed to improve the lot of people, getting them into better jobs.
There was some truth to that, according to a study out of M.I.T., New Frontiers: The Origins and Content of New Work, 1940-2018. David Autor, Caroline Chin, Anna Salomons, and Bryan Seegmiller were able to compare detailed job categories from the U.S. Census Bureau microdata and patent-based measures of occupations’ exposure to “labor-augmenting and labor-automating innovations.”
From 1940 to 1980, new technology coexisted with increased creation of new work in middle-paid production and clerical occupations. There were more jobs, though, unfortunately, many people put out of old occupations didn’t have the background to take the new jobs.
Things changed in 1980 and going forward. First, there was a schism in jobs, with some offering thriving careers for high-paid professionals and others getting low-paid service work. “Innovations that automate tasks or reduce occupational demand slow new work emergence.” The amount of work created by new technologies has been less than the work eliminated. Since 1980, new technology hasn’t created more jobs than it has destroyed.
White collar jobs were being taken before the ChatGPT large language model through what was called automation, but which has included mechanics and artificial intelligence for a long time. Blue-collar jobs have been on the slab for decades longer. People who lose jobs find fewer places to shift careers. This has combined with a weakening labor market.
A new problem has arrived. Many CEOs have pushed AI projects as ways to increase productivity and cut labor costs, especially with hype about generative AI. Boards have been enthusiastic, wanting to know when the benefits would roll in. However, a separate MIT study, The GenAI Divide: State of AI in Business 2025, said a review of over 300 publicly disclosed AI initiatives, structured interviews with representatives from 52 organizations, and survey responses from 153 senior leaders from four major industry conferences, showed that 95% of corporate AI projects had zero return.
High failure numbers in corporate software projects aren’t something new, but in the past it has been on the order of 70%. Now? Only 5% of pilot projects went into full implementation with measurable value.
Organization leaders have little understanding of technology. They focus on pitched projects for sales and marketing because the return on investment seems easy to measure. What actually pays off are areas like procurement, finance, and operations.
And then, employees at 90% of companies already use existing tools like ChatGPT, while only 40% of companies have purchased licenses, putting them in potential legal risk.
Unless companies can get better control, the newest versions of AI may encounter the same challenges that other technologies have faced. Executives and boards will get tired and move on and the promise of cutting headcount could slow and shrink when the people who sign the checks no longer trust the promises.
