Between 2016 and 2024, the number of human-intensive tasks in US jobs actually increased. That is the central finding from MIT Sloan researchers Roberto Rigobon and Isabella Loaiza-Saa, and it runs directly counter to the "AI will take your job" panic that has dominated headlines for the past two years.
Their paper, "The EPOCH of AI: Human-Machine Complementarities at Work," analyzed task structures within occupations using O*NET data (the US government's database of job requirements and worker skills). The key distinction: they looked at what people actually do within jobs, not just job titles, and found that newly added tasks in 2024 required more human capabilities than tasks that were removed.
The EPOCH Framework
The researchers propose five categories of human capability that AI consistently struggles with:
- Empathy and Emotional Intelligence - reading a room, managing grief, building trust
- Presence, Networking, and Connectedness - being physically there, building relationships
- Opinion, Judgment, and Ethics - making calls where there is no clear right answer
- Creativity and Imagination - genuine novelty, not pattern recombination
- Hope, Vision, and Leadership - inspiring people, setting direction under uncertainty
Occupations that score high on these EPOCH capabilities include emergency management directors, clinical psychologists, childcare providers, public relations specialists, and film directors. These are not jobs that will be easily automated regardless of how good language models get.
Augmentation, Not Automation
The study draws a sharp line between automation (handing a task entirely to a machine) and augmentation (using AI to make a human worker faster or more effective). Their argument: most of the real-world AI adoption happening right now falls into the augmentation bucket. A marketer using ChatGPT to draft copy faster is augmentation. A developer using Claude Code to write boilerplate is augmentation. Neither replaces the human - both make the human more productive.
This matches what I see with the AI tools we review daily. The tools gaining the most traction are not replacing people. They are eliminating the tedious parts of jobs so people can focus on the parts that require judgment, creativity, and human connection.
That said, the study's framing has limits. "AI complements workers" is true on average, but averages hide real pain. If your job is 80% routine data entry and 20% judgment calls, AI augmentation might mean your employer needs three people instead of ten. The aggregate employment picture can look stable while specific roles get compressed hard.
The practical takeaway for anyone worried about their career: invest in the EPOCH skills. The more your daily work involves empathy, creative judgment, relationship building, and leadership, the more AI becomes a tool that makes you more valuable rather than less.