MIT report ‘95% of generative AI pilots fail to improve profits and losses’

MIT report “95% of generative AI pilots fail to improve profits and losses”

Gartner “30% of laid-off workers will be replaced by AI and re-employed by 2029”

(Seoul = Yonhap News) Reporter Hayoung Kwon = It is predicted that about 30% of workers who lost their jobs due to being replaced during the introduction of artificial intelligence (AI) will be re-employed by 2029.

As initial expectations for AI investment meet the limits of reality and we have entered the so-called 'valley of disillusionment', it is analyzed that indiscriminate workforce reductions led by AI may result in greater re-employment costs in the long term.

According to the '2026 Future Work Hype Cycle' report recently published by global market research firm Gartner on the 24th, 30% of employees who were replaced by AI and laid off are predicted to be rehired by 2029.

This is because the introduction of AI will deplete the talent pipeline and damage the tacit knowledge and critical thinking capabilities accumulated in the organization, which may result in greater costs to re-secure talent with the necessary capabilities.

Gartner warned that 75% of organizations that only link productivity gains achieved through AI to cost reductions could be overtaken by competitors who reinvest in innovation, modernization, and talent development by 2027.

◇ Generative AI, from expectations to ‘valley of disillusionment’

Recently, there is an analysis that the market has entered a 'valley of disillusionment' where investment sentiment is rapidly declining as initial expectations for generative AI do not lead to actual results.

In fact, the Massachusetts Institute of Technology (MIT) NANDA initiative analyzed more than 300 corporate AI projects and found that 95% of generative AI pilot projects did not produce substantial profit and loss (P&L) improvement effects. Even with corporate investments worth 30 to 40 billion dollars, only 5% achieved results.

In particular, as AI agents that autonomously set goals and perform multi-step tasks are emerging as a core automation tool for companies, concerns are growing that if the human supervision system is not sufficient, unpredictable actions may lead to irreversible results.

‘AI washing’, which promotes the inclusion of AI in marketing even though actual AI functions are missing or limited, is also pointed out as a factor in increasing the gap between expectations of technology and reality.

In a survey conducted by Gartner this year on CEOs and executives, 88% of CEOs responded that they would increase investment in AI, and 71% of boards of directors ordered their companies to make bolder investments in technology.

As the pressure to invest in AI increases, there is an increasing number of cases where suppliers sell existing products or services with only the name AI, or companies invest budgets without verifying actual performance.

◇ Collaboration structure should be designed rather than “replacing people with AI”

Hasty workforce reduction is also considered a pitfall of AI introduction.

Executives in non-IT functions, such as finance, human resources, purchasing, marketing, and sales, said that an average of 18% of their company's employee roles have already been significantly redesigned due to AI.

Experts say that if management who views AI only as a cost-cutting tool undertakes excessive restructuring based on this, productivity may increase in the short term, but in the long term, it may reduce the organization's competitiveness.

The diagnosis is that if the so-called 'AI toolmate', which recognizes AI as a visible collaboration partner in the work flow, is established, a virtuous cycle is possible by lowering employees' resistance and distrust and increasing productivity, participation, and adaptation to organizational culture.

Gartner advised, “Successful companies will use AI to enhance the judgment, creativity, leadership, and decision-making of their employees, not replace it,” and added, “The goal is not maximum autonomy for AI, but to optimally design human-AI dynamics.”