Employee use and perceived impacts on their competence may be behind the slow AI adoption in the workplace

Sixty-eight percent of business leaders polled for the Q4 2024 KPMG AI Quarterly Pulse Survey¹ were planning to invest between $50-$250 million in GenAI over the next 12 months. Considering the investments companies are making in AI, shouldn’t the adoption rate be much higher? Apparently not. A new study from a team of researchers associated with M.I.T.’s Media Lab, covered in a New Yorker article, reports, “(d)espite $30-$40 billion in enterprise investment into GenAI … 95% of organizations are getting zero return.’”  

A significant challenge for companies is the low rate of employee adoption of the technology. The above-mentioned KPMG survey revealed that 46% of business leaders rate slow employee adoption of AI as a significant anticipated challenge of deploying AI to their workers.² The extent of the AI adoption challenge is evident in KPMG’s Q1 2025 KPMG AI Quarterly Pulse Survey, which reported a low pickup rate of 35% (up from 25% in Q4 2024). A New York Times article also reports low adoption at JPMorgan, where about 200,000 of the company’s employees have access to an AI assistant, but only about half of the workers use it. In another study, Competence Penalty Is a Barrier to the Adoption of New Technology, the authors examined barriers to AI adoption among 28,696 software engineers at a “leading global technology company.” The company had installed an AI tool that they had internally shown could improve developer productivity by 30% and actively promoted its use. However, company-wide adoption was low: only 41% of software engineers ever used it during the first year after it was introduced. 

The authors of the Competence Penalty Is a Barrier to the Adoption of New Technology paper hypothesized that individuals using AI at work may be perceived as less competent than those not using the technology, thus being subject to a “competence penalty.” Breaking their results down further, the researchers found that after 12 months, there was a gender gap in usage of 12% between male and female users (43% for males and 31% for females). There was also a 5% gap between mature-age and young engineers after 12 months, with younger engineers adopting AI at a higher rate. They suggest that the low AI use rates of these groups may be due to the competence penalty, and that the penalty “could be particularly pronounced on individuals whose competence is already under more scrutiny, such as females in male-dominated sectors like STEM and mature-age engineers in fast-paced technology environments … .” 

This study involved 1,026 engineers who evaluated the same Python code they were led to believe was written by either a male or female engineer, with or without AI assistance. They then rated the quality of the code and the coder's competence. They found that the supposed adoption of AI reduced the competence ratings of engineers and that female engineers were rated as less competent than their male counterparts. 

With the above findings in hand, the researchers conducted a third study, surveying 919 engineers to see whether software engineers anticipated the competence penalty and thus avoided using AI. The results showed that a higher anticipated competence penalty led to lower rates of AI adoption. The authors concluded that “…anticipated competence penalty explained adoption gaps better than other factors related to AI adoption in the context, such as perceived learning cost.” Moreover, further investigation revealed that “(f)emale engineers anticipated more competence penalty than their male counterparts.” 

The researchers point out a paradox in their findings: “… the very technologies designed to enhance performance can become a liability for the technology user, resulting in a competence penalty and discouraging technology adoption.” They advise companies looking to integrate AI into their workplaces that “(f)raming technology adoption as extending, rather than replacing, human capability can mitigate the competence penalty.” They also suggest that “shifting the focus of employee evaluation from perceived competence to actual work quality may help mitigate competence penalty and improve adoption of technology.” 

 1. The KPMG Quarterly Pulse Survey captures the perspectives of 100 U.S.-based C-suite and business leaders representing organizations with annual revenues of $1 billion or more.

2. Employee adoption was third highest in their findings. The quality of organizational data was the lead anticipated challenge to AI strategies in 2025 (85%), followed by data privacy and cybersecurity (71%).