We May Be Using AI for the Wrong Cognitive Function

How AI lowered the psychological cost of sustained coherence

Before we get into this, don’t you worry. This is not going to be another AI-generated LinkedIn post about “leveraging AI synergies in the age of transformation.” Humanity has suffered enough.

No, this is going to be a true story about how I experimented with Copilot and found a side effect I never dreamed of, but in hindsight, probably should have, because it is so blatantly obvious. And as usual, we don’t see the forest for the trees. Myself included.

A few weeks ago I started using AI not to do things for me, but to help me hold complexity without collapsing under it while burning out.

And before anyone starts foaming at the mouth about the death of human cognition, let me be clear: no, I do not think AI is conscious, omniscient, or the impending arrival of digital Jesus.

LLMs are inherently flawed. They hallucinate. They smooth over uncertainty. They can sound confident while being spectacularly wrong. They inherit biases. They can absolutely become dangerous in the wrong hands. At the same time, they are also spectacular at pattern recognition.

All along my little experiment I kept thinking: We may be using AI for the wrong cognitive function entirely.

Most organizations currently approach AI as an optimization tool. How do we make people faster? How do we automate? How do we reduce headcount? How do we summarize more meetings nobody should have attended in the first place? How do we generate more output into a world already drowning in AI content?

But the moment I started using AI differently, something strange happened.

I got calmer. Not less intelligent. Not more dependent. Less stressed.

And that may be one of the most important things I have experienced professionally in years.

The psychological cost of coherence

See, I’m an Organizational Change Manager. Which means my job is literally to prevent projects from running into walls. Raising risks early, thinking about things we routinely ignore, calling out bullshit, stepping on toes, rocking the boat, detecting incoherence, predicting possible failure.

I am constantly juggling human behavior, leadership dynamics, communication, governance, politics, incentives, psychology, and whatever fresh existential crisis Harvard Business Review decided to summon from the abyss.

Which means my nervous system spends most of its life trying to simultaneously hold:

• structural complexity

• emotional complexity

• political complexity

• narrative complexity

• human fragility

• organizational incoherence

• hidden incentives

• unstated fears

• impossible timelines

• and approximately 47 contradictory project artifacts

And I wish that would be my only problem. I started noticing that any time I raised a difficult question, the reactions ranged from “oh dear” to “Apologies, I need to drop, let’s take this offline.” I literally am the person pulling the “wrong” log from the Jenga tower. Imagine my therapy bill and how popular that makes me in the sandbox.

And then it hit me: The actual exhaustion was not coming from “the work.” It was coming from coherence maintenance. From continuously trying to:

• translate uncomfortable truths into survivable language

• keep track of contradictions

• remember who promised what three months ago and then contradicted themselves

• identify hidden assumptions

• anticipate defensiveness

• package reality carefully enough so the room wouldn’t emotionally reject it before hearing it

In other words, my nervous system was spending enormous energy not merely seeing patterns, but making those patterns communicable inside systems optimized to resist discomfort and measure success in go-live dates.

The moment AI became a mirror

Then I changed one thing. Instead of asking AI: “Do this for me.” I started asking: “Where are the gaps?”

And holy shit. That tiny shift changed everything. Because the moment AI stopped being a vending machine and started becoming a reflective surface, the interaction transformed completely. So I started using it for reflection deliberately. Not: “write the email.” But:

• “show me the gaps in my reasoning.”

• "scan xyz artefact and show me where this can fail”

• “where are the contradictions?”

• “what assumptions are hidden here?”

• “what risks are implied but unstated?”

• “where will accountability drift?”

• “what conditions for success are missing?”

• “what behavior changes are silently assumed?”

• “where is the narrative incoherent?”

• “where can this project plan or strategy fail from a transformation perspective?”

And suddenly the cognitive burden shifted. Not because I outsourced thinking to AI. I remain fully responsible for judgment, ethics, interpretation, and decision-making.

But suddenly I no longer had to manually hold every moving variable simultaneously inside my own head while also emotionally regulating myself and the room around me. AI became an externalized coherence support. Or put differently: AI lowered the psychological cost of sustained coherence. That sentence may sound abstract, but I mean it very literally. I became: less reactive, less irritated and less emotionally flooded. Also more articulate, more precise, more creative and more structurally consistent

AI did not make me smarter. It reduced my cognitive fragmentation. Especially because I no longer needed to spend hours thinking: “How do I raise this without offending everyone and alienating the project team?” I became more neutral.

Are all my projects succeeding now? Of course not. But once you can rely on complex data and cross-reference it with accessible knowledge about organizational behavior, suddenly nobody needs to be the hero or the villain.

From productivity tool to reflective infrastructure

And this is where things started getting very interesting organizationally. Because once I experienced this shift personally, I started wondering: What happens if organizations stop using AI primarily to generate output and start using it to reveal distortion?

Instead of building “agents” that produce more content, I started thinking about agents that expose hidden organizational gaps:

• expectation drift

• governance confusion

• ownership diffusion

• adoption impossibilities

• conflicting assumptions

• psychological safety risks

• recurring friction patterns

• incoherent success metrics

And I know this can sound like a policing mechanism, but think of it more like a pressure test. For example:

What if an AI agent scanned solution designs and identified where organizations assume behavior change without enablement?

What if it detected when accountability was being pushed downward onto people who don’t actually control the levers?

What if it surfaced recurring anxieties before escalation?

What if it highlighted where leadership narratives contradicted operational reality?

What if it continuously revealed where systems were structurally asking humans to perform impossible balancing acts while pretending the issue was “mindset” or “resistance”?

Why organizations struggle to reflect

Before you think “but humans should already be doing that,” yes. They should.

But corporations generate complexity densities of such scale that no human can actually hold them fully. Just show me one person who truly understands how their organization actually works.

A corporation is essentially thousands of nervous systems, incentive structures, political interests, fears, timelines, narratives, and fragmented memories vibrating simultaneously inside one giant socially acceptable panic attack.

Faster dysfunction is not maturtity

Humans cope with this through simplification: hierarchy, slogans, denial, local optimization, scapegoating, buzzwords, performative alignment, “strategic priorities”

and deeply cursed town-halls with inspirational background music.

What AI potentially offers is not perfection. Perfection is a fantasy. Humans are not perfect either, despite centuries of aggressively confident narratives suggesting otherwise. What AI may offer is enhanced reflective capacity. A way to increase an organization’s ability to: detect contradictions earlier, maintain memory continuity, identify systemic patterns, reduce distortion and metabolize complexity without fragmenting

And frankly, I suspect this matters far more than most current “AI transformation” efforts.

Why?

Because if we do know one thing for certain, it is that organizations survive when they adapt and autocorrect. And they can only do that if reflection happens systematically and in practice, not just in theory.

So unless you are sitting on top of infinite reserves and have the luxury of being complacent for a while longer, you absolutely need to understand this:

Your organization will only survive the upcoming years if it becomes mature enough to handle rapid change. And what is the prerequisite of handling change?

Overcoming discomfort.

When honesty doesnt require courage

Most organizations right now are simply accelerating existing dysfunction at machine speed. Faster reporting. Faster slide decks. Faster misalignment. Faster bureaucracy. Faster bullshit.

But faster dysfunction is not evolution.

And only very few organizations have even started considering using AI to free up humans to be more human, which is already an avenue with a lot of merit.

But in my version of AI usage, this goes beyond being “more human.” Because it could expand our tolerance for discomfort and our willingness to tolerate reflection.

Simply because WE no longer need to carry the load of yelling: “The emperor is naked.”

Many organizational failures are psychological failures. The risks were already known. The contradictions were already visible. The assumptions were already fragile. The incentives were already distorted.

But humans avoid discomfort. Systems self-protect. Power resists exposure. Narratives become identity. And accountability dissolves quietly while everyone is busy prioritizing action over meaning.

two possible futures

I’m not saying AI will magically save organizations. In fact, it could absolutely become the opposite:

• surveillance infrastructure

• algorithmic Taylorism

• machine-speed bureaucracy

• “objective” justification for terrible decisions

• industrialized self-deception

That scenario is entirely possible with this avenue. But I increasingly suspect another future is possible too.

One where organizations capable of reflection outperform organizations optimized primarily for narrative management.

One where psychological safety stops being framed as “being nice” and starts being understood as adaptive infrastructure.

One where truth-telling becomes less metabolically punishing because reflective support systems reduce the cognitive and emotional burden of holding complexity alone.

One where leaders stop asking:

“How do we make people adopt AI?”

And start asking:

“How do we use AI to create conditions where better human judgment becomes more sustainable and people are relieved enough to get creative?”

The real question

I genuinely think that is the real question now. The world changed too quickly for our institutional psychology to keep pretending it is still 2010.

• Information asymmetry is collapsing.

• Narratives are leaking.

• Contradictions spread socially at network speed.

• Younger generations grew up inside permanent visibility.

Organizations can no longer rely on coherence theater indefinitely while reality becomes harder and harder to ignore.

And no, I am not naïve enough to think humanity will suddenly become enlightened.

Humans will remain deeply human. We will distort. Exploit. Overreact. Weaponize. Polarize. Monetize. We will probably create an AI-generated corporate mindfulness training narrated by a virtual monk named Zed.

But despite all that, for the first time in a long time, I can actually see a path where technology could increase collective reflective capacity instead of merely increasing efficiency.

A path where we slowly start running out of places to hide.

Just entertain this idea for a moment:

What would happen to your work environment if risks were visible and solved in time?