The Incoherence of Competence

There is a story our industry tells itself about artificial intelligence, and it has the particular comfort of being half true. In this story, AI is a tool — a remarkably capable one, but a tool nonetheless, sitting in the same drawer as the spreadsheet and the slide deck, waiting to be picked up when the tedious part of the job needs doing. It will draft the first version of the brief. It will summarise the market. It will free us, presumably, for the higher work of thinking. We repeat this story in pitch meetings and on conference stages with the confidence of people who have decided, collectively, not to look too closely at what has actually changed. It is a good story. It is also the beginning of the problem.

A nine-hundred-year-old warning

We intend to work only above it.

Nine centuries ago, a jurist and philosopher from Córdoba found himself defending an entire discipline against an accusation that its critics considered devastating. Al-Ghazali had written The Incoherence of the Philosophers, a work of genuine sophistication arguing that rational philosophy — the whole apparatus inherited from Aristotle and refined by generations of Islamic scholars — could never reach the truths that mattered most, and had, in its confidence, produced conclusions riddled with contradiction. Averroes did not respond by softening philosophy's claims or retreating into humility. He wrote The Incoherence of the Incoherence, a point-by-point dismantling of the critique itself, arguing that the confusion Al-Ghazali had located in philosophy was, on closer inspection, a confusion in his own reading of it. The response to being told your tool is insufficient, Averroes understood, is never to abandon the tool. It is to ask, with more rigour than your critic, what the tool was actually for.

The floor has risen. That is not a threat. It is a correction, arriving nine hundred years late, to an industry that had grown far too comfortable mistaking the average for the exceptional.

Above the floor

What the machine actually exposed

We find ourselves in a structurally identical position, and almost nobody in our industry is willing to say so plainly. The tools of brand strategy — positioning, narrative architecture, the long, deliberate work of making an organisation legible to the people it needs to matter to — are being challenged, but not by an argument. They are being challenged by a machine that can produce a plausible imitation of our output in the time it takes to read this sentence. The temptation, exactly as it was for the philosophers of the twelfth century, is to defend our discipline by pointing at what the machine cannot yet do. This is comforting. It is also, if we are honest with ourselves, beside the point.

What survives, and what we believe becomes rarer and more valuable with every model release, is something closer to what Averroes practised nine centuries ago: the discipline of returning to premises rather than defending conclusions, the willingness to sit inside a contradiction long enough to see where it actually originates, and the courage to say the uncomfortable version of the truth before the comfortable one, even when the comfortable one would be easier to sell. A machine can generate the comfortable version in seconds. It has no reason, and no capacity, to generate the version that costs the client something to hear.

Because here is what has actually happened, and it deserves to be stated without the usual hedging. For the better part of two decades, a significant share of the value our industry sold was synthesis dressed as insight. We gathered market data, benchmarked five competitors, distilled it into three strategic pillars and a tone-of-voice document, and we called this strategy because it required real skill to produce quickly and package convincingly. It was not nothing. But it was, in the language Averroes might have used, a conclusion mistaken for a foundation. And synthesis — however well dressed — turns out to be exactly the kind of task a large language model performs with unsettling fluency. Not perfectly. Fluently. Which, in a market that rewards speed and plausibility, is often close enough.

What survives

This is the incoherence we have to name, and it is ours, not the machine's. We continued to charge premium rates for work that was, underneath the formatting, an average of what already existed — and we called it proprietary thinking. The moment a tool can generate that same average in four seconds, for the price of a subscription, the entire premise collapses. Not because the machine is intelligent. It is not, in any sense that matters here. It correlates. It predicts the statistically likely next sentence with extraordinary skill, and in doing so it exposes, with brutal efficiency, every piece of our craft that was never really craft — only competence, repeated often enough to look like expertise.

We do not think this should frighten anyone who came into this work for the right reasons. It should relieve them. For years, the market tolerated frameworks applied like stencils, purpose statements interchangeable across entire sectors, decks that impressed in the room and evaporated by the following Monday. None of that required particular courage. It required competence, confidence, and a client willing not to ask too many follow-up questions. That entire category of work is now, quietly and permanently, worthless — not because a machine stole it, but because a machine revealed that it was never worth very much to begin with.

The difference between competence and judgement

What makes this moment harder than anything we have faced before is not that the machine might eventually match this too. It is that it has already, permanently, removed the cover competence used to provide. There is nowhere left for an adequate idea to hide. A positioning statement that could apply to four competitors in the same category is no longer a mildly disappointing deliverable — it is functionally indistinguishable from something a client could have generated themselves, over lunch, for nothing. The floor of our industry has risen so quickly that what used to pass as a strong presentation now reads, to anyone paying attention, as the bare minimum. And a profession whose entire economic model was built on selling slightly-better-than-average has no room left to stand.

Some of the work we are proudest of inside this agency makes the distinction clear, precisely because it refused the average. When a pan-African financial institution came to us convinced its problem was external perception, the honest diagnosis sat somewhere the client had not thought to look: a fracture between what the organisation lived internally and what it claimed externally, felt as sharply by a relationship manager in a regional branch as by any market analyst. No model, however capable, arrives at that fracture by summarising the brief. It arrives by sitting in a room long enough to notice what nobody quite says out loud. When we advised a performance car brand to deliberately close the door on sixty percent of its own leads, the recommendation ran directly against every optimisation instinct a machine would default to — because desire, unlike conversion, does not respond to more access. It responds to less. An algorithm trained to maximise reach will never recommend scarcity. A strategist who has watched exclusivity build loyalty for twenty years will recommend nothing else.

There is nowhere left for adequacy to hide

This is where the defensive argument our industry reaches for — we are human, and humans cannot be replaced — starts to sound less like a conviction and more like a hope. Being human is not, in itself, a value proposition. It becomes one only when it produces something a prediction engine structurally cannot: a judgment that goes against the probable, a willingness to tell a chief executive something they have organised their entire year around not hearing, an instinct that arrives not from the data in the brief but from the sentence a client left unfinished in a meeting eight months ago. That is not competence. That is something closer to nerve, and it has never scaled, and it has never been available on demand, and it is precisely the thing the last decade of our industry quietly allowed to atrophy while competence did the talking.