AI has made work faster almost everywhere. But, are many organisations confusing sheer speed for actual organisational intelligence?

Dashboards multiply. Reports created in minutes. Output explodes. All while decisions appear to arrive significantly earlier than they ever used to.

Yet, if you have a frank discussion with leadership teams today, you likely find that confidence hasn’t risen in proportion with speed. And, in many cases, confidence might actually be quietly declining.

This pattern is oddly consistent: more activity, but less orientation. More answers, but weaker judgement.

The adoption of AI in organisations may be accelerating work itself, but without reliably making it better. It may even be causing a drift into the trap of conflating outputs with outcomes.

Now, this isn’t because the technology is immature or that the wrong tools/models were selected by an organisation.

Maybe, it’s because the technology is being applied inside systems of decision-making and coordination that were designed in a different era: One of information scarcity, not cognitive overload.

Intelligence doesn’t scale by default

Historically, many organisations have treated intelligence as something you add to the mix:

• More data
• Better tools
• Smarter models
• Faster pipelines

The misplaced assumption here is that judgement and decisions in an organisation will naturally improve once the inputs are strong enough.

However, in practice today, we see that the opposite is happening: Identical AI systems producing wildly different outcomes across teams and organisations.

Some teams become calmer, more decisive, more aligned. Yet, others become noisier, faster, and significantly more brittle.

The difference isn’t the model or tool, but the environment in which thinking actually happens.

Think of intelligence like water flowing through pipes… If the pipes are messy, leaky, or badly connected, adding more water does not help. You just get more mess.

Decades of deep research into tools‑for‑thought and extended cognition already makes this explicit: external systems don’t just support thinking, they participate in it.

Our tools actively shape what can be perceived, compared, questioned, and decided.

So, AI doesn’t replace thinking. But it can stretch it:

• Strong thinking may indeed become stronger.
• But weak thinking can become much louder.
• All while clear judgement is needed more than ever.

So the real question we should be asking isn’t “Are we using AI?”, instead we should ask “What kind of thinking does our organisation actually need?”

This wouldn’t be the first time organisations had to deliberately design their conditions for thinking, but it may be the first time so many are being forced to do it simultaneously.

From tools to cognitive infrastructure

Through my work with the strategic futures studio UN/DO, I have met many senior leadership teams in the last year frustrated with the under-sized impact of their strategic AI investments so far, while feeling forced to continue pushing forward due to anxieties about falling behind the curve.

As a result of these conversations we’ve come to describe this commonly missing perspective in organisations as their cognitive infrastructure.

Now, that might seem like a big term, so let’s break it down…

Cognitive infrastructure is the set of conditions that shape how thinking actually happens across an organisation — not inside one person’s head, but across people, tools, norms, and decision practices.

We can understand it as an organisation’s collective system of thinking.

And similar to roads, pipes, or electrical grids: it doesn’t decide where you actually go, but where you could go. It shapes what’s possible, what scales, and importantly, where things break.

For example, we could visualise a group of colleagues bringing together a combination of skills, tools, norms, expectations, and mental models to turn daily data into clear decisions, and then act.

We can see that AI reliably makes work faster, if not more productive. But judgement — the ability to decide well under uncertainty — depends on how information is interpreted, weighed, and contextualised, not just how quickly it’s produced.

Notice the next time you’re flooded with information.
Does clarity emerge — or just urgency?
Sometimes the problem isn’t that you need better answers,
but to receive them in a space designed for better thinking

Looking at an organisation, cognitive infrastructure is what determines how reliably and quickly:

• information becomes understanding
• understanding translates to decisions
• decisions turn into coordinated action

And for leaders or organisations, the altitude you look at this from really matters here.

Cognitive infrastructure operates at a more fundamental level than culture, technology, or process — it shapes how people interpret information and make sense of situations before behaviour, tools, or workflows even come into play. It sits upstream of:

• Culture — which shapes behaviour, but not necessarily cognition
• Technology — which provides capability, but not judgement
• Process — which sequences action, but rarely embeds reasoning

When organisations struggle to consistently find signal in growing noise, the issue isn’t a missing tool or capability — it’s that the way thinking itself is structured hasn’t yet been deliberately designed. And, we can address this as a metadesign problem.

Instead of deploying more/better/faster tools, we must instead design the organisational conditions for cognition itself so that high‑quality thinking can reliably emerge in these new operating models. In practice, we can understand this in four observable design dimensions: ambiance, scaffolding, amplification & abundance.

So, when AI is simply transplanted into an organisation’s existing systems of thinking, we see activity accelerates often without orientation and diminishing returns.

However, when AI is embedded within a deliberately designed cognitive infrastructure, we see that intelligence can actually compound into greater outcomes than possible before.

The four dimensions of cognitive infrastructure

Through our applied work with leadership teams, a pattern emerged: differences in the organisational impact of AI weren’t random, but mapped to four dimensions of cognitive design within an organisation.

So, how do we climb this pyramid? Let’s work our way from the ground up.

1. Cognitive Ambiance

Cognitive Ambiance is the expectation set by an organisation’s leaders that meaningful thinking with AI is normal, safe, and expected. This is the baseline condition.

In low‑ambiance organisations, we often find that employees may be quietly experimenting with AI in their daily tasks, often cautiously, performatively, or even defensively.

In contrast, those in high‑ambiance organisations don’t ask whether they’re allowed to use AI, they instead ask how they can think better with it.

Ambiance isn’t created by policy. It’s created by what leaders visibly rely on, incentivise, and reference when the decisions matter. Not just citing just the adoption of AI itself, but the highlighting how it actually contributed to meaningful and elevated thinking in the room.

2. Cognitive Scaffolding

Cognitive Scaffolding means designing human-AI interactions and workflows so that an organisation’s purpose, values, assumptions, and trade-offs are surfaced automatically during the process of thinking. That is before decisions are made, not after.

But why? AI tools optimise their responses to meet the explicitly stated objectives of users, instead of responding to the actual intent behind a user prompt or factoring in implicit context of that individual, team or function.

Without scaffolding, organisations increase speed while quietly eroding overall alignment and trust over time. Yes, making decisions may feel faster, but seem increasingly hollow or groundless.

Scaffolding makes key values to an organisation operational at the point of cognition — meaning values show up while decisions are being formed, not just when they’re reviewed.

Effective scaffolding needs clear purpose, decision architectures, and boundary constraints designed-in to our interactions, so that our thinking-tools always surface relevant assumptions and trade‑offs to users before action is taken as part of their natural response, rather than just creating misplaced confidence in an answer or analysis.

At its heart, this is an embedded governance problem rather than a tooling one.

3. Cognitive Amplification

Cognitive Amplification refers to collective fluency in working with AI to enhance judgement and action.

AI amplifies whoever is using it, across the board.

When fluency is concentrated in a handful of star individuals, returns can become spiky and fragile. This group start to produce outsized results, while the organisation becomes increasingly dependent on them.

High‑performing organisations treat human-AI thinking as a systemic capability to be nurtured. They invest in shared literacy, continuous improvement, and common mental models across the workforce so intelligence scales beyond individuals.

Fundamentally, without factoring in the human-side of the equation, even the best tools have little impact in making decisions.

4. Cognitive Abundance

Cognitive Abundance is the final emergent state where an organisation can sustain surplus sense‑making beyond any individual human capability (aka where 1+1 = 3).

In cognitively abundant organisational systems we see:

• Insights compound rather than dissolve
• Decision quality improves even as complexity increases
• The organisation feels calmer, not busier

And this abundance doesn’t mean knowing everything all the time. Instead, it’s resilience: the ability for a team or organisation to remain oriented under uncertainty and ambiguity, without getting lost in the noise.

Cognitive Abundance becomes sustained when ambiance, scaffolding and amplification are singing in harmony. Where clear decisions and judgement simply cut through the noise.

This is a leadership failure (not a tooling one)

None of these dimensions can be fully delegated to a single function. If your AI strategy lives in IT, innovation labs, or pilot teams, you don’t have a strategy (though you may have some plausible deniability!)

Cognitive Infrastructure reframes AI capabilities from a pure technology deployment question, to one of design accountability. It means that the issue for leaders is no longer “Are we using AI?” but “Is our system of human-AI thinking coherent?”

Neglecting any one of these dimensions rarely causes immediate failure. But it can produce significant, but subtle, fragmentation:

• Faster decisions that are less aligned
• Insights don’t compound across teams
• Intelligence that can’t be trusted at scale

The financial risk of doing so is often indirect and delayed. Human capital is absorbed by activity that never becomes a real and repeatable capability for the enterprise.

So, the strategic opportunity is upstream of deployment: deliberately designing how thinking works across an organisation.

Designing for judgement, not just speed

AI is becoming a general‑purpose tool for organisational reasoning.

As it does, competitive advantage will belong less to those who execute fastest, and more to those who remain oriented under complexity.

In the end, returns on intelligence are not accidental. They will be structurally determined, by organisations who can consistently sustain better thinking as the volume increases.

The question isn’t what AI will enable organisations to do next, but what kinds of futures their current systems of thinking are already making inevitable.

Design the cognitive infrastructure and judgement compounds. Ignore it, and speed just makes the noise exponentially louder.

Optional detour

There’s a song from the early 80s that captures this feeling better than most management language ever could.

It’s not about technology. Or work. Or AI. It’s about the strange moment when everything seems normal… but orientation quietly slips.

You don’t need to listen to it to understand this piece. But if you do, notice what it makes you feel before you try to explain it.

This piece draws on the Cognitive Infrastructure framework by UN/DO.

The framework is open‑source under a Creative Commons BY‑NC‑SA 4.0 licence and may be freely used, adapted, and extended for non‑commercial purposes with attribution. Commercial use requires a separate licence. The canonical version lives at weundo.co/cognitive‑infrastructure.

References:

• Clark, A. & Chalmers, D. (1998). The Extended Mind. Analysis, 58, 10–23.
• Clark, A. (2002). Mind, Brains and Tools. In H. Clapin (Ed.), Philosophy of Mental Representation. Oxford.
• Dalladay-Simpson, J. (2008). Tools-for-thought. In Networks of Design, Annual Conference of Design History Society. Universal Publishers.
• Jones, J. C. (1970). Design Methods: Seeds of Human Futures. Wiley-Interscience.
• Latour, B. (2005). Reassembling the Social. Oxford University Press.

Today’s organisations don’t have an AI problem — they have a thinking problem was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.