As AI systems evolve, there’s been a surge of attention and anxiety around how they are reshaping the workforce. The conversation often centers on which jobs are at risk of automation and what machines are capable of. While these are important issues, in our new paper, we offer a different perspective, one that puts humans, not machines, at the center of conversations about the future of work.
To begin, we make a simple claim: Metrics matter.
The way we all measure things not only shapes what we understand, but also how we act. Metrics frame our questions, guide our policy, influence investment, and ultimately shape the future we build. In the context of AI and labor, commonly used metrics like “automation” and “exposure” have limitations. The first focuses solely on what machines might replace. The second combines two different elements – automation and augmentation – and can’t clearly tell us whether AI will substitute for or augment the worker. That distinction matters. Automation shifts tasks from human hands to machines so they can be done more efficiently, with higher quality or simply more cheaply. Augmentation, on the other hand, enhances human workers, giving them new capabilities.
So, to shift the focus from machines to humans, and better frame our collective questions about the future of work, we set out to build better metrics. To do so, we started by asking the question: What are the human capabilities that AI doesn’t easily replicate? The answer would show us where humans can continue to add value, not in spite of AI but when working alongside it.
This question led us to develop the EPOCH framework. EPOCH is an acronym for Empathy, Presence, Opinion, Creativity, and Hope. These and many other related capabilities (see our paper for a complete list) are qualities that remain stubbornly human. From this framework, we derived three metrics to capture tasks and occupations in which AI was more likely to complement humans, instead of replacing them.
We evaluated these metrics across the entire U.S. labor force at both the task and occupation levels. Here’s what we found.
In our non-causal regression models, we found that between 2016 and 2024, there was a measurable shift toward more human-intensive tasks. This shift stemmed from two dynamics: (1) the new tasks introduced in 2024 had higher EPOCH scores than existing tasks, and (2) the frequency with which workers engaged in EPOCH tasks increased.
At the occupational level, the number of jobs with higher EPOCH scores has grown. In fact, all individual-capability categories in EPOCH are associated with positive employment growth over this period. The strongest association is found in occupations that score high in Hope, Vision, and Leadership.
How does this apply in the financial sector?
AI has already had a huge effect in the financial sector. AI-driven models already assess credit risk and detect fraud, while algorithms execute trades in financial markets at higher speeds and volumes than human traders do. However, there are important parts of the financial sector that need and will continue to need human workers. For example:
Financial inclusion: which ensures that underserved communities have access to affordable financial services and is crucial for promoting economic equality and development. AI algorithms perform poorly in this area because there is very little data about individuals who have been excluded from the financial system, leading to reliance on insufficient statistics.
Consumer Experience: Financial services are services. Trust is developed through repeated interaction, but also through human connection. It is not the same to ask a chatbot as it is to ask a person. Here, we suggest that we can learn from Klarna’s experience about going AI-first.
Ultimately, banking requires trust, and trust is something that is built through repeated, consistent, and transparent interactions. Trust requires emotional safety. Thus, while machines can offer consistent interactions, they lack the transparency and emotional ability to inspire trust.
Why does this matter?
Jobs don’t just play an economic role in our lives. They are a cornerstone of our society. They provide identity, status, purpose, and community. So when we talk about how AI is changing work, we are (or maybe should) be talking about how it is changing the human experience.
We believe that by focusing on human capabilities, not just machine capabilities, we can build a more accurate, more meaningful picture of the AI transformation. In doing so, we can shape the future of work in ways that benefit many more than just a few.
The future of work isn’t just about developing new technical skills. It’s about cultivating the capabilities that make us better at working with machines and, more importantly, better at working with each other.
This post comes to us from Isabella Loaiza and Roberto Rigobon at MIT’s Sloan School of Management. It is based on their recent papers, “The EPOCH of AI: Human-Machine Complementarities at Work,” available here, and, “The Limits of AI in Financial Services,” available here.