MIT Proved What the Best Leader I Ever Worked With Already Knew

The best leader I ever worked with opened his first team meeting with a story about a key maker in London. There were maybe twenty of us on the call, scattered across time zones. He pulled up a slide deck that wasn't in company branding. He was slightly jittery, more serious than I'd ever seen him. And then he spent 45 minutes sharing his vision for our work and telling us what he felt he owed us before he ever asked for anything in return.

None of us had experienced that before. A leader showing up with his commitments to us before demanding ours to him.

But that clarity didn't carry through the whole org. Within a few months things started breaking down.

Looking back, there's a pattern from that time that connects to everything I'm about to say. I routinely asked my manager for critical feedback, for someone to tell me what they saw that I couldn't. The best she could give me was "I think you're doing great." Which seems positive, but I wanted to grow. I needed someone to point me toward my next edge. She couldn't do it. Maybe she didn't have the language. Maybe she didn't care enough to look. Either way, I got nothing.

But the problems were bigger than mine. The whole team was struggling, and he picked up on the friction before we even brought it to him. By the time we did, he could see how much we'd been carrying trying to fix it ourselves. His response was to fly us all to London.

Before the trip, he had each of us take a Birkman assessment* and then walked us through our results individually. He wasn't throwing a personality test at us. He was building a shared language for how we each operated before we ever sat down in the same room.

Covent Garden

We landed in London and gathered in a conference room in Covent Garden. The windows looked out onto the kind of streetscape that reminds you with every glance that you're somewhere else. Old stone, narrow roads, architecture that looks like a painting you'd stop in front of at a museum.

Damn near immediately he set the expectation: we need to be vulnerable with each other if this is going to work. But he didn't just ask for it. He modeled it. Every question he posed to us, he answered himself first.

One of the first was about a time you screwed something up at work. Any job. He started with a time he embarrassed himself in an important meeting with huge clients in Japan. Then we went around. Our manager talked about accidentally calling a customer the wrong name. One time. In passing. It was low stakes and everyone in the room felt it. He told her gently that it was the first question, he got it, but to think of something more vulnerable and they'd come back. The rest of us went deeper. I talked about a client who knocked me down a few pegs when I was too impressed with myself as a baby consultant.

He wanted to humanize us to each other first. Then we got into the Birkman results. He'd chosen seven dimensions to deep dive into, things like how you handle emotion, how you make decisions, how you show up when your needs aren't being met. For each one, we plotted ourselves on charts across three layers: how we usually present, what we actually need, and what it looks like when those needs go unmet.

We left London aligned and energized. It was the most thoughtful team-building experience I've ever been part of.

Four months later, he left. The leadership team above him was also challenging. He was the only member not based in North America, and the pull on his time and severe lack of boundary respect became untenable. The system above him couldn't match what he was giving us, so he moved on. After that, the team crumbled.

The thing that held that team together wasn't captured in any system that organization had. No system in that organization could see it, measure it, or protect it. And when it walked out the door, nobody could even articulate what they'd lost.

The invisible gap

I've thought about that experience a lot since then. It was formative, sure, but it exposed something I keep seeing everywhere. The capabilities that actually hold teams together, that make people want to follow someone, that turn a dysfunctional group into an aligned one, those capabilities are invisible to the systems we use to evaluate people. We don't have language for them. We don't measure them. We call them "soft skills" and move on.

And it's not just a hiring problem. It's a talent management problem. It's a retention problem. It's a promotion problem. The same gap that kept my manager from giving me useful feedback is the same gap that let him walk out the door. Other leaders in the org who had worked with him before knew exactly what the company lost. But nobody in his direct leadership structure could see it. The people with the power to protect him didn't have the visibility to know what they had.

MIT asked the right question

About six months after the AI-is-going-to-replace-us-all hype kicked off, I was laid off. I had time to think. Everything I'd experienced with that team, the rough patch, the manager who couldn't tell me what she saw, the system that couldn't protect its best leader, it all kept coming back. I knew intuitively that the capabilities everyone was panicking about losing to AI were the ones we'd never been able to see in the first place. But intuition is a starting point, not a destination. I like to show up with data and receipts, and at the time it felt like nobody had actually studied this.

So my co-founder and I started building a company based on that intuition anyway. The irony of going all in on a business because of vibes wasn't lost on me.

About a year after we started, a paper came out of MIT Sloan that proved the economic value of everything we'd been betting on.

The paper is called "The EPOCH of AI: Human-Machine Complementarities at Work." And it's a breath of fresh air. Almost all of the research on AI and work has been asking the same question. What will AI replace? What jobs are at risk? What tasks can be automated?

The MIT researchers asked the opposite question. What human capabilities actually complement AI? Where does AI fall short, and what do humans do in those gaps? And then they built a way to measure it.

EPOCH: MIT's framework for understanding human value

They identified five groups of human capabilities that enable work in areas where machines are limited. They call it EPOCH:

Empathy and Emotional Intelligence
Presence, Networking, and Connectedness
Opinion, Judgment, and Ethics
Creativity and Imagination
Hope, Vision, and Leadership

An important distinction: the researchers are deliberate about calling these capabilities, not skills. Skills are specific and narrow. Capabilities are broader. They're what emerge when multiple skills interact across different contexts. We'll come back to why that matters.

And before anyone writes these off as the latest repackaging of "soft skills," the researchers address that head on. These capabilities are, in their words, "distinct from so-called 'soft skills,' a misleading term suggesting these abilities are easy to develop. In reality, they are often among the most challenging to cultivate and teach."

The data

Then they tested it. They analyzed every occupation in the US labor market using O*NET data, tracked employment trends from 2015 to 2023, examined current hiring data from 2025, and looked at projections through 2034. The results were striking. And they don't stand alone. PwC's 2026 AI Jobs Barometer, which independently analyzed over a billion job postings across 27 countries, found the same directional shift: roles that emphasize human judgment, empathy, and creativity are growing faster and paying more.

New tasks entering the workforce carry significantly higher EPOCH scores than existing tasks. Tasks that disappeared had the lowest scores of all. Human-intensive tasks are being performed more frequently. And at the occupation level, jobs that score higher on EPOCH capabilities experienced stronger employment growth, higher hiring rates, and better projections for the next decade. Occupations with higher automation risk showed the opposite: consistent decline across all three timeframes.

And the single biggest driver of employment growth? It wasn't creativity. It wasn't empathy. It was hope, vision, and leadership.

That deserves a pause. Hope, in the EPOCH framework, isn't optimism. It's the conviction that a different future is possible and the willingness to act on it before the data tells you it's safe to. That's what my director did on that first Zoom call. He walked into a team that didn't know why it existed and said, here's where we're going, and here's what I owe you to get us there. That's hope. And the economy is rewarding it more than any other human capability. (My co-founder and I started a company on that same instinct before any of this data existed. Usually we just call that delusion, but hope sounds better so I'll gladly go with that.)

Where AI actually breaks

The researchers also mapped where AI hits its limits. Not theoretically. Structurally. They identified five frontiers, but the ones that stuck with me most:

AI excels at interpolation but degrades when asked to extrapolate beyond its training boundaries. It struggles to transfer what it learned in one domain to another. Humans do this constantly without thinking about it, just look up how Velcro was invented.

Some outcomes aren't about finding the right answer at all. They're about building connections, relationships, and shared experiences. AI lacks the capacity for genuinely bidirectional relationships.

And some of the most important decisions in history were made in direct opposition to what the data suggested. Civil rights. Women's rights. Movements driven by a conviction that the current situation was unjust, even when the numbers said otherwise.

The blind spot

That last point, about decisions that defy the data, leads to something the researchers call out explicitly. Algorithmic fairness lies at the heart of AI's limitations. When a group has been systematically excluded from opportunity, the data reflects that exclusion. Not their potential. There's no empirical evidence in a biased dataset to predict what someone could do if they'd actually been given the chance.

Think about resume gaps. A person who spent two years caregiving, navigating a health crisis, or rebuilding after a layoff has been developing EPOCH capabilities the entire time. Empathy. Judgment. Navigating ambiguity without a playbook. But an AI screening tool trained on historical hiring data doesn't see development. It sees absence. And it scores accordingly.

And this measurement gap doesn't stop at hiring. It runs through the entire talent lifecycle. The same systems that can't see these capabilities in candidates can't see them in the people already inside the organization.

This is where the Peter Principle lives. We keep promoting people into roles that demand empathy, judgment, and vision based on metrics that measure none of those things.

The scope of this problem is getting bigger, not smaller. PwC's 2026 AI Jobs Barometer, which analyzed over a billion job postings across 27 countries, found that AI is expanding EPOCH-type requirements across the entire org chart. Entry-level roles in AI-exposed industries are now seven times more likely to require traditionally senior-level capabilities like judgment and leadership. They're calling it "seniorisation." The capabilities that used to only matter at the top now matter everywhere, because AI is absorbing the routine tasks that used to fill those roles.

Which means we need to be able to see these capabilities at every level. And right now, we can't.

Capabilities aren't skills. That matters.

One more thing the researchers are clear about: these are capabilities, not skills. That distinction matters. Skills are specific and narrow. You can name them, observe them, train them. Capabilities are what emerge when certain skills combine, and the combination is different for every person. One person's "hope, vision, and leadership" might be built from strategic foresight and storytelling and initiative. Another person's might be built from empathy and pattern recognition and navigating ambiguity. Same capability. Completely different architecture underneath.

Which means you can't develop these capabilities by training to them directly. You can't run a workshop called "get better at hope." You have to understand the component skills underneath, and those are different for every person.

The spreadsheet

For years, the people who understood the value of these capabilities had no way to prove it. It was intuition. It was "you had to be there." It was the kind of thing you could feel in a room but couldn't put in a spreadsheet.

Well, MIT put it in a spreadsheet. And the numbers say what a lot of us already knew: the most valuable thing in your organization is the thing you've never had a way to measure. That's not a soft problem. It's an economic one. And it's solvable.

So we're building it

We didn't read a paper and see an opportunity. We lived it. We felt it. My co-founder and I both went through layoffs and discovered the same pattern in ourselves: we'd been undervaluing our own contributions for years because we didn't have language for what we actually did. The things that made us effective weren't on our resumes. They weren't in our performance reviews. And there was no credible mechanism for getting them there.

We're building tools that make these capabilities visible. Not through self-assessment, not through a quiz that tells you which archetype you are. Through conversation. You tell the story of a project you worked on, a challenge you navigated, a team dynamic you held together. Our AI listens for the behavioral patterns underneath and maps them against a taxonomy of 52 skills across seven categories. What comes back is a profile built on evidence from your own experience, not a personality label. Because the way people describe their work contains proof of exactly the capabilities the EPOCH research says matter most. You just need something that knows how to listen for it.

*Birkman is the only personality assessment I'll stand behind. It's not gamified. It doesn't reduce you to one of 16 archetypes or tell you which Disney character you are. MBTI, Enneagram, StrengthsFinder, all of them... I have thoughts. If you want to hear them, come find me.

Further reading

• The EPOCH of AI: Human-Machine Complementarities at Work — the full MIT Sloan paper
• PwC 2026 Global AI Jobs Barometer — the billion-job-posting analysis
• MIT Sloan: These human capabilities complement AI's shortcomings — accessible summary of the EPOCH research
• Try Mirra — see what your stories reveal about how you operate