The AI Colleague: What Happens When Your Teammate Isn't Human

There’s a moment in every team meeting where someone says something that changes the trajectory of the conversation. Not the big strategic revelation. Not the data point that shifts the room. The smaller thing. The offhand comment that reframes a problem. The question that nobody was asking but everyone needed to hear. The pause before someone says, “I might be wrong, but…”

That moment doesn’t happen with AI.

I’ve been thinking about this a lot recently, not as a technology question but as a leadership one. Because agentic AI, the kind that doesn’t just answer questions but takes actions, makes decisions, and operates with a degree of autonomy, is no longer a tool you use. It is a teammate you work alongside. And that changes everything about how we understand teams, motivation, and the human experience of working with other humans.

The conversation about AI in the workplace has been dominated by two camps. The optimists see efficiency gains, productivity boosts, and the liberation of human potential from drudge work. The pessimists see displacement, deskilling, and the erosion of meaningful employment. Both camps are focused on the technology. Neither is paying enough attention to the psychology.

Because the question isn’t whether AI will change how teams work. It already has. The question is what happens to the humans in those teams when one of their colleagues has no psychological needs at all.

How Humans and AI Think Differently

The most obvious place to start is with how humans and AI process the world, because the difference is fundamental and it shapes everything that follows.

Dual Process Theory, one of the seven pillars of the STAR Framework, describes two systems of cognition. System 1 is fast, automatic, and intuitive. It operates below conscious awareness, processing patterns, reading social cues, making snap judgments based on experience and emotional signals. System 2 is slow, deliberate, and analytical. It breaks problems into steps, evaluates evidence, and arrives at reasoned conclusions.

Humans operate with both systems, though the balance varies by individual and context. AI operates almost exclusively in System 2 territory. It processes data, identifies patterns, evaluates options, and generates outputs based on logical analysis. It does not have intuitions. It does not read a room. It does not feel the weight of a decision.

This creates a cognitive division of labour that is, on the surface, ideal. The AI handles the heavy analytical lifting. The human contributes the contextual judgment, the emotional intelligence, the ability to sense when something doesn’t feel right even when the data says it should.

But here’s the problem. In most organisations, the division isn’t conscious. Nobody sits down and says, “You handle the analysis, I’ll handle the intuition.” The AI just starts doing things, and the human adjusts. And the direction of that adjustment is almost always the same: the human cedes analytical territory and retreats into oversight.

The cognitive muscle that was exercised by doing the work begins to atrophy from supervising the work. And the team that once operated across both systems becomes a team that operates primarily in one.

For leaders, the question is not whether this division of labour is efficient. It is. The question is whether it is sustainable.

The Three Needs Under Threat

Self-Determination Theory, another of STAR’s pillars, identifies three fundamental psychological needs that drive human motivation: autonomy, competence, and relatedness. When these needs are met, people thrive. When they are thwarted, people disengage. And an AI colleague threatens all three.

Autonomy

Autonomy is the need to feel that your actions originate from within, that you have genuine choice and volition in how you do your work. When an AI colleague is making recommendations, generating options, and in some cases taking actions autonomously, the human’s sense of choice can erode. Not because the AI is controlling them, but because the AI is doing the choosing.

This is subtle. The human still has the final say. They can override the AI. They can choose a different path. But when the AI’s recommendation is consistently good, consistently faster, and consistently more data-informed than the human’s own analysis, the act of overriding starts to feel irrational. The human begins to defer, not because they’ve lost autonomy, but because exercising it feels like a worse decision.

Over time, this creates what SDT researchers call an “external locus of causality.” The human begins to experience their work as driven by the AI’s outputs rather than their own judgment. The motivation shifts from intrinsic to extrinsic. And that shift is corrosive.

Competence

Competence is the need to feel effective, to experience mastery and growth in what you do. When an AI colleague can do parts of your job better than you can, and faster, and without fatigue, the experience of competence is directly challenged.

This is not a hypothetical concern. Doctors using AI diagnostic tools report feeling less confident in their own clinical judgment, even when the AI improves outcomes. Analysts using AI-powered modelling tools report feeling less ownership of the insights they produce. The tool makes them better at their job and worse at feeling good about their job.

The Big Five personality model, another STAR pillar, helps explain why this effect varies across individuals. People high in conscientiousness, those who derive satisfaction from careful, methodical work, may find the AI’s speed and efficiency threatening to their sense of craftsmanship. People high in openness may adapt more readily, seeing the AI as a creative partner rather than a competence rival.

Relatedness

Relatedness is the need to feel connected to others, to experience belonging and mutual recognition in your work. This is the need most obviously threatened by an AI colleague, because an AI cannot relate. It can simulate conversation. It can generate responses that feel empathetic. But it does not experience connection, and the human interacting with it knows this.

A team is not just a collection of individuals working towards a shared goal. It is a social group with norms, rituals, shared language, and mutual recognition. When one member of that group is not human, the social dynamics change. The inside jokes don’t land the same way. The moment of shared frustration after a setback doesn’t bond the same way. The celebratory drink after a win doesn’t include the AI.

This is not sentimentality. It is psychology. The human need for relatedness is not a soft benefit that can be optimised away. It is a fundamental driver of motivation, engagement, and performance. And an AI colleague, no matter how capable, cannot meet it.

The Motivational Erosion

What emerges from this analysis is a pattern I’d call motivational erosion. Not a dramatic collapse. Not a sudden disengagement. A gradual, almost imperceptible decline in the internal motivation that drives people to do their best work.

The mechanism is straightforward. SDT tells us that when autonomy, competence, and relatedness are thwarted, intrinsic motivation declines. The result is a team that looks productive on the surface but is hollowing out underneath. The AI is producing outputs. The humans are reviewing and approving them. The metrics look good. But the humans are less engaged, less challenged, and less connected to the work and to each other than they were before the AI joined the team.

This is not an argument against AI. It is an argument for understanding what AI does to the humans it works alongside, and for designing the human-AI partnership with the same psychological rigour that we’d apply to designing any other team structure.

The Automation Bias Trap

There is a specific risk in the human-AI partnership that leaders need to understand: automation bias. This is the tendency to over-trust automated systems, to accept their outputs uncritically, and to suppress one’s own judgment in favour of the machine’s recommendation.

Automation bias is not a sign of laziness or stupidity. It is a cognitive shortcut that makes sense in most contexts. When a system is consistently right, trusting it is efficient. But it creates a vulnerability. When the AI is wrong, the human who has learned to trust it may not catch the error, because the cognitive habit of critical evaluation has been replaced by the cognitive habit of acceptance.

This is particularly dangerous in high-stakes environments. A medical team that defers to an AI diagnostic may miss the case where the pattern recognition fails. A financial team that relies on AI modelling may not question the assumption that produces the outlier prediction. A marketing team that follows AI-generated strategy may not notice when the algorithm’s model of the customer diverges from reality.

The defence is not to distrust AI. It is to maintain the human capacity for distrust, to keep the critical thinking muscle active even when the AI is doing the heavy lifting. This requires deliberate practice: regular exercises where the team evaluates the AI’s outputs critically, protocols that require human sign-off at key decision points, and a culture that rewards questioning the machine, not just approving it.

What the AI Cannot Feel

Appraisal Theory of Emotion, another STAR pillar, describes how we evaluate events and situations for their relevance to our goals and wellbeing. The appraisal process generates emotional responses that inform our behaviour: anxiety when a threat is detected, satisfaction when a goal is achieved, frustration when progress is blocked.

An AI colleague does not appraise. It does not feel. It processes data and generates outputs, but it has no emotional stake in the outcome.

The absence of emotional stake changes the team dynamic in ways that are easy to overlook. When a project hits a setback, the humans in the team experience frustration, disappointment, maybe even fear. The AI processes the new data and adjusts its model. When a project succeeds, the humans experience satisfaction, pride, maybe even joy. The AI updates its parameters and moves to the next task.

The emotional dimension of teamwork is not a nice-to-have. It is the mechanism through which teams process experience, build resilience, and develop the shared understanding that allows them to perform under pressure. An AI colleague cannot contribute to this process, and its presence can dilute it if the team begins to calibrate its emotional responses to the AI’s emotional flatness.

The Leadership Question

So what does good leadership look like in a team that includes AI colleagues?

It starts with understanding that the AI is not just a tool. It is a presence that reshapes the motivational and cognitive architecture of the team. The leader’s job is not to manage the AI. The AI manages itself. The leader’s job is to manage the humans alongside the AI, ensuring that the three fundamental needs, autonomy, competence, and relatedness, are met even as the nature of the work changes.

This means protecting the human space. Creating moments for genuine connection that don’t involve the AI. Preserving opportunities for deep analytical work that exercise critical thinking, even when the AI could do it faster. Explicitly recognising the human contribution that the AI cannot replicate: the contextual judgment, the emotional intelligence, the creative leap that no algorithm can predict.

It means understanding that different team members will respond to the AI differently. The Big Five gives us a starting point: people high in openness will adapt more readily, while those high in conscientiousness may struggle with the disruption to their process. Leaders who see these patterns can manage the transition with precision rather than hope.

And it means accepting a fundamental truth that the optimists and pessimists both miss. AI does not replace human psychology. It sits inside it. It changes the conditions under which human motivation, cognition, and emotion operate. And those conditions are the leader’s responsibility.

The Moment That Matters

Let me return to where I started. The moment in a team meeting where someone says something that changes the trajectory of the conversation. The offhand comment. The unexpected question. The pause before “I might be wrong, but…”

That moment is the product of psychological safety, cognitive engagement, and human connection. It is the product of a team that trusts each other enough to disagree, that cares enough about the work to think deeply, and that feels connected enough to take a risk in front of each other.

An AI colleague cannot produce that moment. But a good leader can protect the conditions that make it possible, even when the AI is in the room.

Especially when the AI is in the room.

The STAR Framework synthesises seven psychological theories into four primary mindsets and twelve distinct archetypes. It is the foundation of the STAR Operating System, an award-winning model for understanding human behaviour in organisational, consumer, and leadership contexts. David Chadderton is the creator of the STAR Framework and author of The STAR Framework: Rewriting the Rules of Consumer Engagement (NYC Big Book Award, 2025), The STAR Operating System, and Dear Algorithm, It’s Not Me, It’s You. He is CMO at Homes for Students, VervLife, and Orla.