There is a principle embedded in economics that rarely fails: over time, and all else being equal, the price of any service tends toward the cost of delivering it. Competition erodes margin. Commoditisation follows standardisation. The only durable defence against this dynamic is scarcity – either of skill, knowledge, or access.
For much of the twentieth century, that scarcity was structural. The cost of delivering complex analytical, advisory, or creative work was high precisely because it required years of training, specialised tools, and significant human effort. That structural scarcity kept prices elevated and margins intact.
Artificial intelligence is dismantling it rapidly.
The Cost Curve Has Shifted
For an expanding set of services – drafting, summarising, coding, modelling, designing – the marginal cost of production has fallen close to zero. A report that once took a day can be produced in minutes. Analysis that required a team can be prompted by an individual. The gap between the cost of conception and the cost of execution has collapsed.
This does not mean the work is being done poorly. In many cases, AI-assisted outputs are indistinguishable from – and in some dimensions better than – those produced through purely human effort. The quality floor has risen even as the cost floor has dropped.
For practitioners in any field where this dynamic is playing out, the implications are significant.
Two Responses to a Falling Cost Floor
There are two constructive responses to this shift.
The first is to develop intellectual property that cannot be easily replicated. Proprietary frameworks, datasets, methodologies, and insights grounded in deep domain expertise and lived experience represent genuine scarcity. AI can assist in their production and communication, but it cannot substitute for the underlying judgment that gives them value. A well-constructed framework for evaluating retirement solutions – one that captures the trade-offs inherent in managing longevity, market, and liquidity risk – is valuable not because it is hard to typeset, but because it reflects years of accumulated thinking that cannot be prompted into existence.
The second is to stay at the bleeding edge of technological advancement. If the cost of delivery is falling, the advantage lies in being among the first to harness new capabilities – before they become table stakes. Practitioners who develop AI fluency early compound their lead: they produce more, explore wider, and iterate faster. Those who wait until adoption is universal find themselves with neither the IP advantage nor the efficiency gain.
The Productive Tension
These two responses are not in conflict – they are complementary. Deep IP anchors the value you offer. Technological fluency determines how efficiently you can create and deliver it.
The risk is in assuming that one is sufficient without the other. Accumulated expertise without technological leverage becomes increasingly expensive to deliver relative to AI-augmented competitors. Technological fluency without underlying insight produces work that is generic, derivative, and – eventually – indistinguishable from the outputs of anyone else with access to the same tools.
The price of a service will always converge to the cost of delivery. The question is whether what you deliver carries a form of scarcity that no model can easily replicate – or whether it is, at its core, a commodity waiting to be automated away.