Reflection has a reputation problem in organisations.

It is often associated with delay, indecision, or a kind of managerial indulgence that feels disconnected from the pressures of real work. When people hear calls to slow down, to think more carefully, or to examine assumptions, what they often hear is a threat to momentum. Something that will get in the way of delivery.

While that reaction is understandable; it is also based on a misunderstanding. Paralysis does not come from reflection; it comes from uncertainty that has nowhere to go.

Subscribe now

In many organisations, uncertainty is treated as a problem to be eliminated as quickly as possible. Decisions are expected to resolve ambiguity rather than work with it. Questions are tolerated only insofar as they can be answered cleanly. The result is not clarity, but compression. Complexity is pushed downwards, outwards, or sideways, where it reappears later as friction, failure, or fatigue.

This is the environment in which reflection becomes dangerous, not because it is slow, but because it is unsupported. When people are asked to reflect without being given permission to act differently as a result, reflection turns inward. It loops and becomes self-referential, so it feels unproductive because it is unproductive. Nothing in the system can move in response to what is being noticed.

That is not reflection, it is containment, and the alternative is not speed, it is direction.

Share

Reflection that enables movement is oriented toward purpose rather than certainty. It does not ask people to analyse everything. It asks them to notice specific things that matter for the work at hand. It creates just enough pause to adjust course before acceleration locks the wrong pattern in place.

This distinction matters in the context of AI because AI systems reward momentum. They make it easier to move from question to output without friction. In doing so, they reduce the natural pauses in which judgment used to occur. Reflection, in this context, is not a philosophical luxury. It is a compensatory mechanism and a way of reintroducing deliberateness into an environment optimised for flow.

The fear is that if we stop to reflect, we will never start again. The evidence suggests the opposite. Organisations that fail with AI tend to move too quickly at the beginning, locking in assumptions they have not examined, and then slow down dramatically later when consequences emerge. What looks like decisiveness early on often produces paralysis downstream. Reflection, done well, shifts effort upstream.

It does not ask for exhaustive analysis, it asks for bounded but timely questioning. It creates space to ask whether the work being automated is actually understood, whether the decision logic is explicit, whether the outcomes being optimised for are the ones the organisation claims to value.

Reflection without paralysis has edges. It is attached to real work, not abstract improvement and it happens close to decisions that matter. It is time limited but not rushed, so it produces direction rather than answers. This is where metacognition becomes practical.

When individuals and teams are able to notice when they are being carried by fluency rather than understanding, they can intervene early. When they are able to name uncertainty without being penalised for it, they can keep moving while staying aligned. When disagreement is treated as signal rather than obstruction, reflection sharpens action rather than delaying it.

Organisations that manage this well do not reflect more, they reflect differently. They do not open everything up at once. They choose specific domains where judgment is heavy and consequences are real. They allow reflection to inform design choices before those choices harden into systems. They accept that some questions do not need to be answered immediately but do need to be held consciously.

This is particularly important with AI, because once systems scale, they resist revision. What could have been questioned becomes infrastructure, and what could have been adjusted becomes policy, so reflection delayed becomes rework. The opposite of paralysis, then, is not speed, it is responsiveness.

Responsiveness requires the capacity to sense when conditions are changing, to interpret what that change means, and to adjust course without drama. Reflection is what enables that sensing, and action is what tests it. The two are not in opposition, in fact they are coupled.

If reflection feels like paralysis in an organisation, it is usually because the organisation has not built pathways for insight to translate into action. People are invited to notice, but not to change. They are asked to think, but not to decide. Over time, they learn that reflection is unsafe, and they stop doing it altogether.

AI will not fix that; it will in fact expose it.

The work, then, is not to choose between reflection and momentum, but to build the conditions under which reflection can inform movement without overwhelming it. To treat reflective capacity not as a brake, but as a steering mechanism. This is not slower work, it is earlier work and it is the only way to avoid the far more expensive paralysis that arrives later, when systems fail in public and organisations scramble to explain decisions they never really examined in the first place.

The final question is not whether reflection will slow you down. It is whether you prefer to pause while you still have choices, or later, when those choices have already been made for you.

The Art of Unravelling is offered as a gift, sustained by those who feel called to support the weaving. If you’d like to help tend the fabric of this work, you can contribute via the link below.

Buy me a coffee

There is a great deal of conversation about skills in the AI era. We all need to learn prompting, be AI literate and fluent in how the different models work. We value this technical understanding and change readiness in our people and all of these matter, however none of them touch the core issue. What is missing is not skill in the usual sense, it is a capacity. Specifically, the capacity to notice how we are thinking while we are thinking, and to adjust in response to that noticing, which is known as metacognition.

Subscribe now

Metacognition is not new and it is not exotic. It sits beneath good judgment, sound decision making, and the ability to operate under uncertainty without collapsing into certainty too early. It is what allows a person, or a group, to pause and ask whether the frame they are using still fits the situation they are in.

In the context of AI, this capacity becomes critical.

As we have already seen, AI systems do not simply produce outputs. They shape the cognitive environment in which decisions are made. They influence what feels salient, what feels complete, and what feels unnecessary to revisit. In that environment, the ability to monitor one’s own reasoning is no longer a nice to have. It is a safety mechanism.

Without it, people drift. However, at first this drift does not look dramatic. It looks like competence, moving smoothly from question to answer with confidence and it looks like work getting done. What disappears is the moment of self-interrogation, the quiet check that asks whether the answer is grounded, whether the assumptions still hold, whether the situation has shifted in ways the system cannot see.

Share

Research into professional use of generative AI consistently shows this pattern. When people actively plan, monitor, and adjust their thinking, AI can improve outcomes. When they do not, performance stagnates or declines, even among highly capable professionals. The difference is not intelligence...it is metacognitive engagement.

This is why so many AI initiatives feel simultaneously impressive and fragile. The outputs are fluent, the systems appear to work and yet something essential feels thinner, less robust and more easily knocked off course. That something is reflective capacity.

What makes this particularly challenging is that metacognition does not announce itself. You cannot mandate it through policy. You cannot install it through software. You cannot replace it with governance language. It shows up in how conversations are held, how disagreement is tolerated, how uncertainty is handled, and how quickly people feel pressured to resolve ambiguity.

In this sense, AI transformation is inseparable from organisational culture, not because culture is a soft backdrop, but because culture determines whether reflective pauses are possible at all.

This is where the connection to foresight and futures becomes unavoidable. Good futures work does not begin with prediction; it begins with noticing. Noticing which assumptions are being treated as fixed, which trajectories are being extrapolated without question, and which possibilities are being excluded because they are uncomfortable, inconvenient, or destabilising to current plans.

Integral foresight rests on the same capacity as metacognitive AI use. The ability to step outside a dominant frame, to hold multiple perspectives at once, and to remain oriented to purpose rather than momentum. Without that capacity, futures work collapses into trend watching, and AI work collapses into acceleration. In both cases, the failure mode is the same, acting more quickly inside an unquestioned worldview.

This is why metacognition is so often misunderstood in organisational settings. It is mistaken for introspection, or hesitation, or lack of decisiveness. In reality, it is what allows decisiveness to remain aligned with reality as conditions change.

Organisations that lack this capacity tend to oscillate between overconfidence and paralysis. They rush forward until something breaks, then slow everything down in response. What they rarely do is build the steady reflective muscle that allows them to move with care without stopping entirely.

AI amplifies this dynamic, so in organisations where reflective capacity is already present, AI can extend it. It can surface patterns, test assumptions, and support more thoughtful sensemaking. In organisations where reflective capacity is thin or suppressed, AI accelerates drift because it rewards fluency over understanding and speed over judgment.

This is why training people to use AI without strengthening metacognition is not just insufficient, it is actively risky. Prompting skills increase output but they do not increase discernment. Technical literacy improves access but not judgment. Without metacognitive capacity, these skills simply make it easier to move quickly in the wrong direction.

The most important work of AI transformation, then, does not sit in the technology roadmap. It sits in the conditions that allow people to notice when they are being carried along by coherence rather than conviction, by momentum rather than meaning.

This is slow work, but it is not optional. It is the work that makes the difference between AI as an accelerant of existing patterns and AI as a companion in rethinking how work is actually done.

If the previous post argued that AI reflects the organisation back to itself, this one names the capacity required to look at that reflection without flinching, and without rushing to either denial or control.

The question that remains is whether organisations are willing to invest in that capacity, or whether they will continue to search for substitutes that feel easier to scale.

The Art of Unravelling is offered as a gift, sustained by those who feel called to support the weaving. If you’d like to help tend the fabric of this work, you can contribute via the link below.

Buy me a coffee

There is a persistent fantasy running through much of the AI conversation, it sounds like this.

If we bring in a sufficiently advanced system, it will somehow lift the organisation beyond its current habits. It will smooth over cultural friction and will compensate for weak decision making. It will clarify what we have not been willing to clarify ourselves.

This fantasy has a long history, it appeared during digital transformation as well, just dressed in different language. Then, the hope was that new systems would force discipline, alignment, and integration. That once the technology was in place, behaviour would follow.

What actually followed was years of adaptation, workaround, and quiet repair. Unfortunately, AI does not break this pattern, it intensifies it. What organisations are discovering now, often with discomfort, is that AI does not improve culture…it reveals it. It also does not correct decision logic, it scales it. More frustratingly, it does not magically resolve ambiguity, instead it reproduces whatever assumptions are already doing the work.

Subscribe now

What you see is what you get

If an organisation is thoughtful, careful, and explicit about how judgment is exercised, AI can extend that care. If an organisation is fragmented, rushed, or performative about its values, AI will reproduce that fragmentation at speed. If an organisation relies on people to quietly compensate for structural problems, then AI will remove that compensation and expose the problem directly.

This is not a failure of the technology; it is a property of it.

AI systems learn from data, but they are shaped by framing. They respond to how problems are described, what is treated as important, and which trade-offs are considered acceptable. Those choices are cultural choices, whether or not they are named as such.

When people say that AI went wrong, what they often mean is that it behaved in a way that made visible something the organisation was already doing, but not acknowledging. Bias that had been distributed across many small decisions becomes explicit, incoherence that had been buffered by human judgment becomes formalised and values that were stated but not enacted become impossible to ignore.

Share

This is why AI feels so confronting in some settings. It does not politely adapt itself to the organisation’s self-image. It reflects the organisation back to itself with uncomfortable fidelity.

This is also why there is no shortcut here. No amount of tooling, governance language, or ethical positioning can substitute for the work of examining how decisions are actually made. There is no configuration setting that compensates for unclear purpose. There is no model that fixes a culture that relies on silence, avoidance, or heroics to function.

Attempts to bypass this reality tend to take on a familiar shape. Organisations copy the surface features of successful AI programs without doing the underlying work. They create centres of excellence, ethics boards, and strategy documents. They adopt the language of responsibility without changing how trade-offs are made. They perform seriousness rather than practicing it. This is cargo cult behaviour, the rituals are there but the outcomes are not.

AI is particularly unforgiving of this kind of pretence because it operates at scale and speed, it removes the margin in which inconsistency can hide. What had previously been softened by discretion becomes stark because what had been distributed across people becomes concentrated in systems.

This is why the work before implementation matters so much. Not preparation as a checklist or a phase gate, but as a genuine examination of how the organisation thinks. How it frames problems, handles uncertainty and decides what matters when values collide. How much discretion people are actually trusted with, and how much is merely assumed.

Without that examination, AI will faithfully reproduce the organisation’s existing logic, including all of its blind spots, while giving the impression of objectivity and progress. With it, AI can become something else maybe not a corrective force, but a revealing one. A way of making decision logic visible, discussable, and therefore changeable.

This is the real offer of this moment, not transformation through technology, but transformation through seeing. Seeing what has been normalised, been carried quietly and where purpose and practice have drifted apart.

There is no magic here. There is no leapfrogging the work of culture, judgment, and accountability. AI will not take you somewhere your organisation is not already prepared to go. It will take you more quickly and more clearly to where you already are.

The choice, then, is not whether to adopt AI, but whether to be honest about what it will reflect back when you do.

The Art of Unravelling is offered as a gift, sustained by those who feel called to support the weaving. If you’d like to help tend the fabric of this work, you can contribute via the link below.

Buy me a coffee

Every organisation tells a story about how its work happens. That story lives in process maps, role descriptions, policies, operating models, and dashboards. It is usually coherent and reassuring. It suggests that if the right steps are followed in the right order, the right outcomes will follow.

That story is not false, but it is incomplete.

Subscribe now

Alongside it exists another version of the organisation, one that rarely appears in formal descriptions. This version lives in practice rather than documentation. It is enacted in moments of hesitation, adjustment, and judgment. It shows up when the situation does not quite fit the category, when the data is technically correct but practically misleading, when following the rule would produce a worse outcome than bending it.

This work is not exceptional, and it is constant.

A clinician notices that a patient meets the criteria for discharge but hesitates because something about the presentation does not sit right. A policy officer realises that two regulations intersect in a way that produces an unintended outcome and quietly negotiates a workaround. A case worker slows a decision because the circumstances do not match the category they have been assigned. A consultant reframes a client question because answering it directly would miss the real problem.

None of this appears in the official process, yet without it, the organisation would not function.

Organisations do not ignore this work because they are careless. They ignore it because it is difficult to formalise and uncomfortable to acknowledge. It resists standardisation. It depends on experience, situational awareness, and moral judgment. It introduces ambiguity into systems designed to eliminate ambiguity.

Over time, this work becomes backgrounded. It is treated as personal style rather than organisational capability; it is assumed rather than examined and it is relied upon without being named.

During the digital transformation years, this invisible work became the shock absorber. As systems multiplied and processes became more rigid, people learned how to compensate. They translated between incompatible tools; they held exceptions in their heads; they maintained parallel records and they developed informal rules about when to trust the system and when to override it.

This was not inefficiency, it was adaptation. It was the organisational equivalent of nitrogen-fixing plants that appear after land is cleared, the species that stabilize soil and create conditions for something more complex to eventually grow. Essential, invisible, and chronically undervalued.

The problem is that adaptive work leaves almost no trace. It does not appear in datasets or logs. It is rarely captured in metrics. When it succeeds, nothing happens yet when it fails, the failure is often attributed to the individual rather than the system.

When AI systems are introduced, they are almost always built on what is visible. The documented process and the recorded data. The categories that survived abstraction and the assumptions that were stable enough to encode. What is missing is the work that made those assumptions survivable.

As a result, AI systems do not encounter the organisation as it actually exists. They encounter a simplified version, stripped of the adjustments that allowed it to cope with reality. The system then performs exactly as designed, and the organisation is surprised by the outcome.

Decisions that once involved hesitation now move straight through. Cases that once triggered informal escalation are processed automatically. Situations that previously relied on tacit judgment are resolved by proxy variables that were never intended to carry that weight.

This is how harm enters quietly. Not because anyone made a reckless choice, and not because the system malfunctioned, but because the work that prevented harm was never acknowledged as work in the first place. When that work is overwritten, nothing signals its absence until the consequences appear downstream.

This is why organisations so often insist they are automating routine work, only to discover later that the routine was being held together by non-routine effort. The so-called edge cases turn out to be common. The exceptions turn out to be the work. The judgment assumed to be incidental turns out to be essential. AI does not create this problem, it simply reveals it.

As they were implemented, digital systems required work to be translated into fixed steps. They demanded stable categories, predefined flows, and unambiguous rules. Anything that did not fit had to be pushed aside or absorbed by people. Over time, organisations contorted themselves to suit their systems. Processes accreted layers, and temporary fixes hardened into permanent structures. The original purpose of the work was often buried under years of accommodation.

AI behaves differently. It does not require the same degree of upfront specification. It does not need every step named in advance, it can work with intent, with context, with descriptions of what matters rather than instructions for how to proceed. It responds not only to processes, but to the mental models people hold about the work itself.

This is the missed opportunity in most AI deployments. Instead of using AI to revisit what the work is for, organisations are using it to speed up how the work currently happens. Instead of asking what outcomes they are trying to produce, they are encoding the paths they already know, including all the compromises and distortions those paths have accumulated over time.

So, we get our old assumptions, newly automated.

However, because AI responds to framing and purpose, it offers another move. It makes it possible to surface and test the mental models that have been shaping work quietly for years. It can expose where processes no longer serve their original intent. It can reveal where rules have drifted away from outcomes. It can make visible the gap between what the organisation says it values and what its systems actually optimise for.

This is not because AI is wise, it is because it is literal.

When people are forced to articulate what they are trying to achieve, what trade-offs they are willing to make, and what conditions matter most, inconsistencies surface. When those articulations are compared across roles and cases, patterns emerge. The organisation begins to see itself not just as a set of workflows, but as a set of beliefs about how work should be done.

Share

That kind of seeing was difficult with digital systems. They were too rigid and too costly to change once built. AI, used reflectively, lowers the cost of inquiry. It allows organisations to ask different questions of their own work before locking anything in.

This is the opportunity that sits alongside the risk. If AI is treated purely as an automation layer, it will entrench everything that has accreted over time. If it is treated as a way of interrogating purpose, judgment, and outcome, it can help clear some of that accumulation. Not by wiping the slate clean, but by making visible what no longer belongs.

That does not happen automatically, it requires restraint. and resisting the urge to encode too quickly. It requires using AI first as a way of thinking with the organisation, rather than acting on its behalf.

To look twice at work, then, is not only to notice what has been invisible. It is to recognise that this moment offers a rare chance to realign how work is understood with what it is meant to do. To ask not just how the process runs, but whether it still serves the purpose it was built for.

That chance will not last forever. Once systems are scaled, they harden. Mental models become infrastructure. What could have been questioned becomes assumed again. The mirror does not force change, it offers clarity. What an organisation does with that clarity is still a choice.

The Art of Unravelling is offered as a gift, sustained by those who feel called to support the weaving. If you’d like to help tend the fabric of this work, you can contribute via the link below.

Buy me a coffee

By the end of 2025, the numbers were everywhere.

Headlines announcing that most AI initiatives were failing to scale. Reports quietly revising early optimism downward. Case studies describing pilots that never moved beyond experimentation, or deployments that produced output without producing trust. The tone shifted from inevitability to unease. By the start of 2026, even the most optimistic enterprise voices were talking about deflation, digestion, and value-realisation rather than acceleration.

For anyone who lived through digital transformation, there was a familiar feeling in the body. Not surprise, exactly, but recognition, because we have been here before.

For more than two decades, organisations have been told that transformation was primarily a technical challenge. If the right systems were implemented, behaviour would follow. If data were centralised, insight would emerge. If processes were standardised, efficiency and coherence would arrive behind them.

Instead, what arrived was complexity.

Enterprise systems multiplied rather than unified. Administrative burden increased even as automation promised relief. Workarounds became permanent. Staff learned how to keep things moving despite systems that did not quite fit the reality of the work. Transformation happened, but not in the way it was promised. By the time AI entered the organisational imagination at scale we were already standing in depleted ground, a graveyard of previous system implementations.

Subscribe now

Introducing a new species into exhausted soil

What makes this moment different is not the presence of failure, but the speed at which it appears. AI surfaces problems within months, not years. Initiatives stall not because the models are weak, but because the organisation cannot absorb what the technology is making visible.

Like an invasive species introduced into a weakened ecosystem, AI doesn’t cause the fragility, it simply exploits and amplifies it. In healthy systems with clear nutrient flows, stable relationships, and resilient structures, new species can integrate. In degraded systems, they take over quickly, outcompeting what was already struggling to survive.

The familiar response follows: stronger governance protocols, better communication, tighter controls, additional reviews. These measures have their place, but they don’t address the substrate problem.

During the earlier wave of digital transformation we didn’t fail because the tools were inadequate. We failed because we treated transformation as something that could be delivered to the organisation, rather than something that had to be metabolised by it. We layered systems on top of unresolved complexity and asked people to cope. We depleted the organisational soil, people’s capacity for attention, adaptation, and collective sensemaking, without allowing time for restoration.

AI removes the coping layer. It doesn’t let organisations pretend the work is simpler than it is, nor does it extract judgment from context without consequence. Where digital systems could be patched, AI amplifies the underlying pattern. Where workarounds could be absorbed by people, AI formalises assumptions and scales them at speed.

What grows in degraded ground

Here’s what’s uncomfortable to acknowledge: in some contexts, the rapid spread of AI might stabilise short-term dysfunction. Like weeds that prevent erosion on bare soil, AI can fill gaps, produce output, and keep things moving when human capacity is exhausted.

But if you’re trying to protect something more vulnerable, maybe careful judgment or contextual discretion, even the quiet intelligence that holds complex systems together, then the same rapid growth becomes a problem. What stabilises in one frame destroys in another. So the question isn’t whether AI is inherently good or bad. The question is: what kind of system are we trying to sustain, and does introducing this species support or undermine it?

Share

Restoring organisational substrate

Standing in this graveyard is not a reason to stop, but it is a reason to look carefully at what went wrong before deciding how to move. If AI is failing to take root in the ways we hoped, or taking root in ways we didn’t intend, it’s worth asking what condition the ground is actually in:

1. Audit what can actually be absorbed: Before scaling any AI project, assess whether your organisation has the substrate to support it. You will need clear decision-making structures, shared understanding of purpose, capacity for ongoing adaptation, and the ability to metabolise what the technology makes visible.

2. Restore capacity where it’s depleted: If the audit reveals exhaustion, your teams stretched thin, judgment operating on autopilot, no space for reflection or course-correction, then invest in restoration first. Training, deliberate learning time, reduced cognitive load, cultural permission to slow down when needed.

3. Create conditions for healthy integration: Bring together the people who understand the work as it’s actually done: frontline staff, domain experts, those who hold tacit knowledge, alongside technical and governance teams. Let them continuously validate AI outputs against real-world context. This creates the feedback loops that allow new capabilities to integrate rather than displace.

The choice now isn’t whether to adopt AI as the market has largely decided that. The choice is whether we repeat the same pattern, introducing new species into degraded ground and hoping for the best, or whether we take the time to understand what kind of ecosystem we’re actually working with.

What thrives in depleted soil is rarely what you intended to cultivate.

The Art of Unravelling is offered as a gift, sustained by those who feel called to support the weaving. If you’d like to help tend the fabric of this work, you can contribute via the link below.

Buy me a coffee

When a new species enters degraded ground, it doesn’t fail because the organism is weak, it fails because the substrate can’t support what it needs to thrive. The same is true when AI enters cognitively depleted organisations.

Subscribe now

There is an assumption running through much of the AI conversation. If capable, well-trained professionals are given access to powerful tools, performance will improve. Intelligence will compound because better inputs will produce better outputs but, in many cases, the opposite is happening.

Highly skilled professionals report feeling more fluent and more productive when using AI. Independent evaluation shows their accuracy declining. Decisions are made more quickly, with greater confidence, and with less scrutiny. Errors don’t announce themselves as errors because they arrive wrapped in plausibility.

This isn’t because smart people suddenly become careless. It’s because AI creates what researchers call a jagged technological frontier. Some tasks are easily accomplished while others, though seemingly similar in difficulty, lie completely outside its current capabilities.

Share

Here’s the critical part. For tasks outside that frontier, professionals using AI perform significantly worse than those working without it, not just no better, actually worse.

This pattern shows up consistently across domains. In the landmark study with Boston Consulting Group, consultants using GPT-4 on tasks within AI’s reliable range saw quality improvements of over 40%. But when the same consultants used AI for tasks requiring contextual judgment, tasks outside the frontier, performance dropped sharply. The problem wasn’t the AI. It was that users couldn’t reliably tell the difference between tasks where AI excelled and tasks where it floundered.

This is cognitive depletion meeting computational confidence.

When organisational attention is already stretched thin, when people are managing too many tools, holding too many workarounds in their heads, compensating for systems that don’t quite work, AI doesn’t restore capacity. It offers a different kind of shortcut and like most shortcuts through depleted ground, it works until it doesn’t.

Traditional automation bias occurred when people over-trusted machines perceived as precise. AI introduces something subtler. Over-trust through coherence. The output sounds thoughtful, it feels complete and it arrives fluently and quickly. The mind, already tired, relaxes. Our attention shifts from evaluation to acceptance. What had previously required active reasoning becomes passive review.

This is where expertise becomes a risk factor rather than a safeguard. Experienced professionals trust their ability to judge quality, but research shows they demonstrate miscalibrated trust, over-relying on AI precisely where it’s weakest and under-using it where it excels. Confidence increases even as calibration degrades so, the work looks faster, and the outputs look polished, but the underlying reasoning becomes thinner.

If the first post in this series established that we’re standing in depleted organisational ground, this is what happens when we introduce AI into that depletion. The system doesn’t restore substrate, it exploits what’s left.

Organisations already stretched by digital transformation, already compensating for fragmented systems, already asking people to do more with less cognitive capacity, now introduce a technology that removes the friction that used to signal slow down. It produces outputs that feel complete before they’ve been properly examined. It rewards momentum over monitoring. It makes cognitive shortcuts feel like efficiency.

This isn’t a warning against AI. It’s a warning about the conditions under which AI is introduced. In organisations where reflective capacity is healthy, where people have space to think, permission to slow down, and support for careful judgment, AI can extend that capability. In organisations where attention is already depleted, AI accelerates drift.

What organisations are discovering, often uncomfortably, is that AI literacy isn’t enough. Prompting skills aren’t enough. Technical training isn’t enough. What’s missing is the capacity to notice when coherence is standing in for correctness. To recognize when fluency is masking incompleteness. To sense when the work requires re-engagement rather than acceleration.

This is metacognitive capacity, otherwise known as the ability to monitor your own thinking while you are thinking, and it’s precisely what gets lost when organisational soil is depleted.

The question isn’t whether smart people can use AI well, of course they can. The question is whether organisations are willing to restore the substrate that makes good judgment possible, or whether they’ll keep introducing new species into exhausted ground and hoping for different results. What thrives in cognitive depletion is rarely what you intended to cultivate.

The Art of Unravelling is offered as a gift, sustained by those who feel called to support the weaving. If you’d like to help tend the fabric of this work, you can contribute via the link below.

Buy me a coffee