When Output Is Automated, What Makes People Valuable?

Thursday 26 March. The train departing from Rotterdam was quite empty at 9:15, the way Dutch trains are when you travel outside the rush hour. I quickly checked the mobile: many of my colleagues based in Europe were already online by then, their green dots lighted up in Microsoft Teams, messages requiring my attention. I went back finishing the article I was reading.

I work fully remote. Most days, my visibility inside the organization is a function of what I produce: PowerPoints shipped, e-mails sent, progress on a document, things that light up dashboards and inboxes. That’s the deal. You’re remote, so you’d better be productive.

But this morning I was on a train to the office. Not for a meeting. Not to work together on a deadline. I went because a friend and former colleague who recently joined the firm organized a lunch together, and I wanted to see him as we barely cross path lately, and rescheduling another time just did not feel quite right. That’s it. We grabbed coffee, went out for a long lunch, talked a bit about work, talked mostly about interesting things that had little to do with work. I ran into other colleagues I never met before in the hallway and at the desk. My friend introduced me to one of them, and we ended up scheduling a conversation that I already know will be valuable. None of this would have happened on a video call. The most productive thing I did all day was stop being productive.

A few days earlier, I’d been reading Ryan Warner’s piece in Psychology Today about the power of being valued beyond productivity. I first read it on my reMarkable tablet, scribbled some notes in the margins, then re-read it during the commute and scribbled more. Warner’s argument is grounded in organizational psychology: self-determination theory, humble leadership, the effects of holistic recognition.

But something kept nagging me. The argument felt incomplete. Not wrong. Just not urgent enough. Because what Warner frames as good leadership practice I think it is about to become something much bigger.

The Research We Already Had

I should say upfront: I’m not a psychologist and do not have a PhD. I’m someone who reads a lot of psychology, and what I’ve been reading recently has started to feel less like “interesting research” and more like a “survival manual”.

The basics are well established. Self-determination theory, developed by Deci and Ryan and validated across decades of research, holds that people need three things to thrive at work: autonomy, competence, and relatedness. A 2026 meta-analysis by Hagger and McAnally Star, covering roughly 190 studies, confirmed that satisfying these needs consistently predicts engagement, job satisfaction, and performance. That’s not a niche finding. It’s one of the most replicated results in organizational psychology.

Then there’s the humble leadership literature. A review of 115 empirical studies by Kelemen and colleagues found that leaders who demonstrate accurate self-assessment, openness to being wrong, and genuine appreciation for others’ strengths produce measurably higher engagement, retention, and psychological safety. Chandler and colleagues confirmed these findings in a quantitative meta-analysis. The leader who says “I don’t know, let’s figure this out together” isn’t being weak. The data says they’re building something structural.

And recognition matters. A study of over 25,000 employees by Jo and Shin found that recognition significantly boosts engagement, while transformational leadership both increases engagement and reduces burnout. Meanwhile, Gallup’s 2025 report shows U.S. employee engagement has sunk to 31%, its lowest point in a decade. The other sixty-nine are showing up, doing the work, and not particularly caring about any of it.

Here’s what strikes me about all this research: it’s been telling us for years that valuing people as whole human beings, not just as output machines, produces better results. But many organizations treated it as a nice addition. A layer of frosting on the productivity cake. The implicit assumption was always that productivity was the real transaction between employee and employer, and everything else was a bonus. Well… I think that assumption is breaking.

Agentic AI Changes the Equation

The machines got good. Not just “faster at spreadsheets” good. Agentic AI systems can now reason, plan, and execute across complex knowledge work: analysis, content generation, code production, data synthesis, research, operational decisions. These aren’t assistants waiting for instructions, but agents capable of autonomous work and getting things done. And… they don’t take coffee breaks. They don’t need encouragement. They don’t have bad Mondays.

And yet, the productivity revolution hasn’t arrived. A 2026 NBER study surveying roughly 6,000 executives across the U.S., U.K., Germany, and Australia found that nearly 90% of firms reported no measurable impact from AI on either employment or productivity over the past three years. Economists are calling it the Solow paradox revisited, after Robert Solow’s famous 1987 observation: “You can see the computer age everywhere but in the productivity statistics.” Forty years later, same sentence, different technology.

But the direction of travel is unmistakable. McKinsey’s research identifies skills rooted in social and emotional intelligence (conflict resolution, design thinking, negotiation, coaching) as the ones that will remain “uniquely human.” AI can’t yet set direction, interpret ambiguity, or make value-driven decisions. Those are human capacities. For now.

So here’s the strategic flip that Warner’s article hinted at but didn’t fully spell out: if AI can replicate what humans produce, then defining human value by what they produce is a losing game. The employee whose value proposition is reducible to output is now competing on territory where machines hold structural advantages in speed, cost, and scalability. That’s not a fair fight. Well, it’s not supposed to be a fight at all.

And that transforms the “value people beyond productivity” argument from a humanistic principle into a strategic imperative. The psychological needs that self-determination theory identified (autonomy, competence, relatedness) aren’t just things that make employees feel good. They describe the dimensions of work where human contribution is irreplaceable. AI doesn’t experience autonomy. It doesn’t form genuine relational bonds. Its competence comes from training data, not from lived experience, judgment under ambiguity, or ethical reasoning shaped by consequence.

The qualities that humble leadership research says build psychological safety (self-awareness, openness to being wrong, appreciation for others’ perspectives) are precisely the qualities that resist automation. An AI system can simulate humility. It can’t model the vulnerability that makes it meaningful.

The Labor Market Already Knew

The labor market has been pricing it in for quite some time.

David Deming’s 2017 analysis in The Quarterly Journal of Economics showed that between 1980 and 2012, jobs requiring both high technical skills and high social skills grew by 7.2 percentage points, with wages increasing 26%. Jobs requiring high technical skills but low social interaction? Well, they actually declined by 3.3 percentage points, and wages grew only 5.9%. The market was sending a signal but maybe not many people were listening.

John Burn-Murdoch’s 2026 analysis in the Financial Times, building on Deming’s framework with updated data, found the divergence has widened further. In the technology sector specifically, employment in high-social-skill roles (developers, systems analysts) roughly doubled from 2001 levels. Employment in low-social-skill roles (programmers, statisticians) barely moved. Wages tracked the same split.

The counterintuitive finding is worth sitting with: proximity to automation does not predict vulnerability. Isolation from human interaction does. Even in the most automation-exposed fields, the roles that combined technical capability with collaboration, stakeholder management, and relational coordination consistently outperformed those built on technical execution alone.

This resonated with me. I work in technology. And the most valuable moments in my work aren’t the outputs I produce per se. They’re the conversations where I help a colleague think through a problem, the trust I build with a client that makes a difficult project possible, an insightful exchange of views with a technology partner, the unplanned coffee with a friend that turns into a new idea.

Going Beyond Transactional

I think we need to start with what “gets measured”. Job architectures built around task completion will erode as AI absorbs those tasks. Performance systems that only measure what was delivered are already losing relevance. What matters more: how the work was done. How others were developed. How trust was built. How ambiguity was navigated without rushing to premature resolution.

Recognition practices carry new weight here. Organizations that only recognize output are inadvertently signaling that employees should compete with AI on AI’s terms. That’s a competition that I think humans will systematically lose. Organizations that recognize judgment, ethical reasoning, creative dissent, mentorship, and the ability to hold complexity are investing in the capabilities that define the human layer of a hybrid organization.

But it’s not just about what organizations do. It’s about what each of us does.

I’m truly convinced we need to work hard at upskilling for the AI age. I do it on a daily basis, and I share my knowledge with family, friends and colleagues as much as I can. Learning to use these tools well, understanding their limitations, developing the judgment to know when AI output is good enough and when it needs a human eye. That work is real and it’s necessary.

And I’m equally convinced that upskilling alone isn’t enough. We need to invest in connection, even when it looks like wasted time. Take the meeting that has no agenda. Pick up the phone instead of sending an email (even worse, an email generated by AI). Go to the office, sometimes, just to be in the same room as someone, consciously exchanging a couple of productive hours in front of the laptop with commuting time (which, by the way, you can turn productive in other ways). Not because it shows up on a productivity dashboard, but because it builds something no AI can build: trust, shared context, and the kind of understanding that makes good judgment possible.

This can happen remotely, too. A thoughtful video call where you actually and actively listen (no multitasking!), where you ask the question you’ve been avoiding, where you sit with silence instead of filling it with slides. Connection isn’t about location but rather about intention.

I keep thinking about younger colleagues, about my own kids. If AI is eventually going to displace many jobs, what skills do they actually need? Not just technical fluency with AI. They need to know how to build trust across a table, how to navigate disagreement without walking away, how to sit with uncertainty instead of rushing toward the first answer and take it for granted. They need to know that being good at something a machine can’t do is more valuable than being great at something a machine can do better.

Coaching them, mentoring them, helping them develop these capacities: that might be the most important work any of us does in the next decade.

The Load-Bearing Structure

This morning, I took a train to see a friend. The carriage was half empty. I didn’t produce anything. I didn’t check off all the tasks I had to. And I came home with more energy, more engagement and more ideas than I’ve had in weeks.

Trust, psychological safety, meaning, belonging, ethical judgment: this was never a soft benefit. It was the load-bearing structure. It still is. Maybe, now that the machines can handle the output, we’ll finally treat it that way.

---

Note: I’m not a psychologist, and this article doesn’t attempt clinical analysis. It’s one technologist’s reading of research that felt suddenly urgent. All claims are linked to their sources. The Gallup engagement figure of 31% reflects one specific measurement framework; other organizations using broader definitions report significantly higher figures.