Augmented Individuals, Diminishing Organisations

At Sentech Africa Tech Week last week, Caitlin Nash, who did an incredible job of hosting the event as well as moderating many of the panels, made a comment that resonated. "Perhaps it is an opportunity to be more human than we ever have been before." 

This was in relation to, as was many of the topics of the event, being the accelerated adoption of AI in organisations. 

The optimism is real. So is the tension underneath it.

Because the trajectory most organisations are on right now is the opposite. Distributed work forces, faster staff turnover, increasing asynchronised work, and now accelerated AI implementation. Last week I wrote about teams "codifying how the business works" into markdown files and prompt libraries. The instinct is right. The architecture may be missing. One of the possible failure modes is that capable individuals become extraordinarily capable on their own, while the team around them doesn't develop with them. They become solo operators with very good tools.

Which sounds fine, until you notice what gets lost.

Jamie Metzl, in Superconvergence, makes a very interesting point. People today are not meaningfully smarter than our ancestors ten thousand years ago. Genetically, the raw intelligence is the same. What separates us is the accumulated, shared, structured knowledge we have inherited and built on. We are not smarter people. We are the beneficiaries of a cultural architecture that holds learning in place, transmits it across generations, and lets each generation compound on the last.

Two great accelerants in our history have compounded the effect of shared knowledge: writing, and later the printing press. Neither replaced human dialogue. They raised its altitude. Each generation could start higher up because someone had bothered to write it down, and then someone else read it, argued with it, taught from it, built on it. The text was the floor. The interaction was where the next layer of understanding got made. And so the cycle repeated itself:

A knowledge flywheel where people used what came before, engaged, built new context, and documented it, ready to distribute for the next engagement, whether themselves or those that come after.

Strip out the text and every conversation starts from zero. Strip out the conversation and the existing written knowledge doesn't compound. 

That layering is the part that matters.

Metzl frames the rate of human innovation as: (number of people) × (education level) × (network exposure) × (tool capacity)

The organisational parallel is exact.

Every organisation has a pool of capable people. But capability alone doesn't determine performance. What determines it is whether the organisation's collective understanding is captured, shared, and made reusable. The choices, the context, the reasoning. Without that, teams fragment. Decisions get relitigated. The same problems get solved twice, then three times, by different people who never see each other's working.

A team of ten brilliant people who keep what they know in private inboxes and individual memory is likely no more effective than ten isolated individuals. A team of ten ordinary people operating from a shared, well-governed knowledge environment will outperform them. Reliably.

The organisational version of Metzl's formula reads something like:

(headcount) × (individual capability) × (quality of shared knowledge) × (ease of access and contribution) = organisational rate of learning and execution

The first two terms are what most organisations hire for. The last two, almost no one has deliberately built. Without them, compounding returns and growth are severely hindered. 

Now layer AI on top. Following the internet, AI is the next great accelerant in this lineage. It is entirely possible, today, for ten ordinary people to become ten augmented people doing the work of forty. Faster, sharper, more polished. By every individual measure, they are delivering.

The problem doesn't show up in their work, but it does show up between them.

If those ten default to solo, augmented mode, each in their own AI-shaped reality, each accumulating their own private context, the network effect quietly collapses. One person solves their version of a problem brilliantly. Someone else solves a slightly different version next week. A third person, the month after. The outputs are polished individually. They don't combine, don't compound, don't make the eleventh case any easier than the first. When someone leaves, the context they were holding goes with them. The team starts again. Brilliantly. From zero.

Over time, the collective output gets less effective, not more. Each person stays sharp. The organisation gets duller. And ultimately less effective.

None of this is, strictly speaking, an AI problem. Organisations have been losing context, repeating themselves, and depending on what travels in people's heads for as long as organisations have existed. AI didn't create the issue, but it has however removed the excuse to ignore it.

Thordur Arnason wrote about this shift into solo-mode a few months ago in a great piece called Ready Player None. His diagnosis is behavioural and sharp: AI is making individuals extraordinarily capable while pulling them out of the interactions where shared understanding actually gets made.

John Cutler of "The Beautiful Mess" newsletter, as well wrote an adjacent point in his article Seeing Everything, Understanding Nothing (The Context Trap), sharper at the epistemological level. Most organisations treat context as a transmittable package: write it down, share it, and assume alignment follows. AI accelerates the temptation, because we now have tools that can ingest enormous volumes of context and hand it over in what appears to be perfect packaging. But not all context works that way. Some can be documented and supplied ahead of time. Often, it can only be produced through the interaction itself. The meeting generates it, the engagement carries it. You cannot transmit your way to alignment when the work requires enactment.

Neither argument dismisses the floor I mention. They sharpen the layering. Documentation that pretends to be the interaction is the failure mode. Documentation that gives the interaction something coherent to stand on, is the point. 

The reason multiplayer mode is "still loading," as Arnason puts it, isn't a design gap or a cultural problem. It's a missing capability.

Most organisations have never had to build the shared knowledge layer that makes multiplayer worth showing up for. Proximity, shared offices, and the way work flowed through teams did the job ambiently, without anyone needing to design it. In a distributed, AI-augmented world, that ambient scaffolding has thinned to almost nothing, and very little has been built to replace it. Not because no one has tried to design it; plenty of systems exist. 

But, because it isn't a system to deploy. It's a capability that needs to be developed, and the capability has two halves that depend on each other. 

• The structure: a knowledge environment that's architected, governed, kept current. 

• The practice: the everyday discipline of contributing, refining, reading, deliberating, building on what's there. 

The structure makes the practice feel worth the effort. The practice keeps the structure trustworthy. Together they compound. Apart they decay.

Right now, most “collaboration” in organisations isn't necessarily producing new understanding. It's recovering old context that should already be structured, findable, and current. Meetings are held to re-establish what was already decided. Calls to re-explain what should already be documented. Working sessions to re-derive what should already be reusable. And these happen without the team realising it.

That isn't interaction creating value. That's transmission failure masquerading as collaboration. 

None of this is lost on the people inside it. They feel the waste, and it's part of why so many quietly default to solo mode, to async, to deep work, to anything that reduces the time spent in another meeting recovering context. They're right about the waste. They're wrong about what's causing it. The problem isn't engagement. It's engagement without a floor.

A well-architected knowledge environment doesn't replace interaction. It clears the waste, so that when people do interact, they're operating at the level where interaction actually produces something new. Bringing forward well-structured previous context, allowing people to leverage their thinking and build new richer context.

Which brings Caitlin's comment back into focus. To be more human than we have ever been before requires the network effect: the compounding that comes from shared understanding, not individual brilliance scaled. The contextual reasoning, the messy generative argument, the judgement that doesn't compress into a prompt, these are the things humans do uniquely well. They only compound when there's something coherent for them to stand on, and a practice of keeping it that way.

That isn't a system anyone can design once and deploy. It's a capability the organisation has to develop and hold in place.

Knowledge architecture, in other words