A controversial survey by Upwork Research Institute found that while 96% of C-suite leaders expect the use of generative AI tools to increase overall productivity levels, 77% of surveyed employees say they have actually decreased their productivity. In fact, the survey contradicts previous research showing a positive correlation.
Upwork Research surveyed 2,500 workers across the US, UK, Australia, and Canada, including C-suite executives (50%), full-time employees (25%), and freelancers (25%).
The picture provided by executives contradicts strikingly what employees report. Indeed, 81% of leaders at companies that deployed AI-based tools believe the overall company productivity increased, compared with only 42% of leaders at companies that did not adopt AI-based tools. Yet, many leaders expect more:
One in two executives at companies using AI believe their company is falling behind their competitors (51%) and that their workforce’s overall productivity levels are stalled due to lack of employee skills and adoption (50%).
Despite their expectations about the benefits of using AI tools, approximately three-quarters of surveyed executives admit they have no training plan in place for their workforce, and only 13% maintain they developed a well-implemented strategy. Instead, AI adoption seems to be emerging bottom-up, thanks to early adopters and innovators in the workforce.
This is somewhat coherent, with 47% of surveyed workers saying "they have no idea how to achieve the productivity gains their employers expect". Interestingly, executives believe the opposite, with 37% of them saying their workforce is highly skilled and comfortable with these tools. The reality is only 17% of employees feel they are.
Surveyed workers list several factors that contribute to their productivity loss:
For example, survey respondents reported that they’re spending more time reviewing or moderating AI-generated content (39%), invest more time learning to use these tools (23%), and are now being asked to do more work (21%).
This leads Upwork researchers to explain these data through the well-known "productivity paradox", referring to the slowdown in US productivity during the 70s and the 80s while IT tech adoption was rapidly climbing.
By deploying new technology—no matter how exciting and full of potential—without updating our organizational systems and models, we risk creating productivity strain [...]. We risk another productivity paradox with generative AI if we don’t fundamentally rethink the way we work.
Among the measures that Upwork suggests to improve AI adoption are leveraging non-traditional talent, co-creating productivity metrics, and moving toward skill-based approaches rather than job roles.
Echoed by Forbes' contributor Bryan Robinson, who writes, "scientific findings released today contradict those expectations and re-ignite apprehensions about AI’s impact on employee overload and burnout", the survey has raised some criticism on the Internet.
The first factor of criticism is the source of the survey itself. Being Upwork, a marketplace for freelancers, several commentators see a credibility problem with a survey that makes freelancers appear more attractive to large organizations. Another reason for concern is the lack of a full explanation of the survey methodology.
Others are more open to accepting the survey as trustworthy, but tend to highlight the limits of AI tools themselves, saying, "AI models (generative or not) are useful in specific cases, not all cases. Failing to acknowledge that and failing to strategise accordingly only leads to short term success and long term pain", or even the lack of understanding of AI tools in management: "The only people excited are people who simply do not know enough to know they are wrong". It goes without saying that many commenters report an increase in their productivity and an actual workload reduction by automating small, repetitive tasks.
As a final note, it seems the majority of academic research on the topic tends to posit a positive correlation between the use of AI tools and productivity, but their analysis falls out of the scope of the present article.