How will AI and emerging technologies impact the US workforce? Since the arrival of ChatGpt in less than three years, such arguments have usually fallen into one of two categories. How AI strengthens the workforce and how AI-driven automation destroys the workforce.
New paper from MIT Sloan Post Doctor Associate Isabella Royza and Professor We take a different approach by asking, “Which human abilities complement the shortcomings of AI?”
This approach shifts arguments from confusion and alternatives to labor to human capabilities. “In future areas, the focus is often on machines rather than humans,” Roysa said. “We wanted to focus on what humans can do. We don't want to instill fear in people's minds. We wanted to show how AI can complement workers.”
To answer their questions, researchers developed a human-intensive competence framework that produced three key indicators:
- Applicable Risk Score
- Potential luxury scores
- With scores, ethics, creativity, and emotional intelligence that show whether the task depends on a particular human abilities that AI lacks.
The pair applied each of these metrics to all US tasks and occupations identified by the US Bureau of Labor Statistics occupational information database O*Net, and defined 19,000 tasks across approximately 950 different jobs.
Their conclusion? Work that relies on human traits such as empathy, judgment, and hope is less likely to be replaced by machines.
AI is influencing advanced skills work
While technological advances have always sparked concern among workers, advances in mechanization, automation and digitalization have generally improved the quality of people's work over time. This may not be the immediate case of artificial intelligence.
“I feel that's different because AI threatens to replace abilities that are deeply connected to cognitive abilities,” writes Loaiza and Rigobon, saying that AI can brainstorm, create content and solve problems.
“AI is moving faster and the impact on the workforce is different,” says Loaiza. “While previous waves of technology tended to have a negative impact on skilled workers, AI has impacted workers regardless of their achievement of education.”
However, AI has limitations. You cannot infer from a small dataset or extrapolate well beyond the training dataset. Issues from two or more viable solutions, as well as decisions based on shared experience, pose challenges for AI as well.
And critically, AI is struggling with subjective beliefs. This is characterized by researchers as decisions based on a variety of results that differ from what the data suggest. “Some of the most transformative decisions in human history have been driven by beliefs that reject the status quo, even if general data appears to support it,” they write the women's suffrage and the civil rights movement, citing when convictions stand before the status quo.
“[Humans sometimes] We make a decision not because the data says it is possible, but from principles, not because we should do it,” Roysa said.
Humans can do things that AI can't do yet
To begin their research, Loaiza and Rigobon outlined the human abilities of five groups, represented by the acronym Epoch.
- Empathy and emotional intelligence. AI may be able to detect emotions, but humans can create meaningful connections and share what a person is experiencing. Occupations such as social work and education are very clear indications.
- Being, networking, and connection. The role of nursing and journalism reflects the importance of physical presence in building connections, promoting innovation, and working with colleagues.
- Opinions, judgments, ethics. Humans can navigate open-ended systems such as legal professions and the science industry, but AI has a hard time grasping concepts such as accountability and responsibility.
- Creativity and imagination. Humor, improvisation, and “visualization of possibilities beyond reality” remain human capabilities, as researchers have said. They are especially valuable in their design and scientific work.
- Hope, vision, leadership. Grit, perseverance, and initiatives further embody the human spirit. This means starting a new company, taking on the challenge despite the long-term potential for success.
Armed with the framework, Roysa and Rigobon studied approximately 19,000 working tasks using O*NET data in the context of automation and augmentation, and in relation to human capabilities.
These so-called task statements are descriptive, but specific to each profession. This means that there is little overlap between descriptions of tasks for different occupations, making it difficult to identify similar tasks distributed among different roles.
To overcome that limitation, Loaiza and Rigobon grouped the tasks into 750 clusters. For example, one cluster consists of similar tasks related to building sales websites for various occupations. Another consists of tasks related to design review in a variety of fields, such as game design, sculpture, and digital imaging.
The researchers then assigned three scores to each cluster of tasks. We assigned automation risk scores, potential continuity scores, and epoch scores. This indicates whether a task is associated with an epoch feature that automates and protects the task. From there, they compared the total changes in employment in the US workforce from 2016 to 2024 with their results.
Researchers found that tasks with high risk of automation and augmentation involve corresponding high-risk unemployment. Automation-related findings reflect AI as an alternative to human labor. With regard to increased growth, the decline in employment is attributed to increased productivity. This means that businesses can produce more without hiring additional workers. Extensions are not necessarily partial automation, Leuza noted. It simply means that workers can complete tasks more quickly than before, or perform tasks that they could not do before.
Leading an AI-led organization
Directly on MIT Sloan
Sign up now
Human abilities make work more resilient
In contrast, all epoch ability groups were related to employment growth, with the biggest impact being the second biggest influence from the ability and ability to view the hopeful human beings. These results write that “the shift towards a more human-intensive job not only by the tasks performed in each occupation, but also by the number of people employed in more human-intensive jobs.”
The positive impact of human abilities was a pleasant surprise to Roysa, but it wasn't necessarily unexpected. “There are many top-level managers in the decision-making role, especially in developed countries,” she said. “There is a lot of value for human workers.”
For Loaiza, the findings reinforce the notion that AI strategies must emphasize that they will increase workers rather than replace workers. It also provides enterprise leaders with a roadmap for high-class workers. It is easy to overlook when training workers for the future caused by AI use, paying particular attention to “the basic qualities of human nature.”
“In many areas, workers cannot be replaced entirely,” Royza said. “If you're looking to be a disruptive innovation or a truly transformative business, people have a big role to play.”
Read next: When humans and ai work best together – and when each one is better on their own
