The vocabulary behind the calls.

One person built the PorcelainSheer catalogue in days — photography, copy, production. A team of three would have taken weeks. The capability compressed.

What did not compress: the judgment about whether it was worth building. When execution becomes cheap, that question gets skipped. Compression removes the friction that used to force it.

In consulting, this gap runs on a reliable cycle. An internal team builds an AI tool. It demonstrates. Then someone asks what it would take to scale it, integrate it, or pass a security review. The answer is another team and a rewrite.

The demo was never a product. Production is a product problem — it requires product thinking before the first line is written, not after the demo lands.

The minimum efficient team for a catalogue, a site, or a working software tool is now one person. This is visible in both directions.

On the personal side, the leverage is real. On the consulting side, the same one-person team is arriving as an independent contractor — and becoming competition for firms whose pricing was calibrated to a team that is becoming optional.

When you remove a layer of people from a working arrangement, you discover how much faster the remaining work moves. The coordination tax was invisible because it looked like how work gets done.

It does not disappear when AI takes over execution. It compresses into artefacts — task packets, prompts, result files. The only tax worth paying is the one that moves a judgment forward, not the one that synchronises context.

When AI drafts the operating model for a large utility client, the consultant has no mechanism to audit the reasoning chain. The output looks correct. The logic is coherent. But coherent and auditable are not the same property.

The client does not hold the model accountable. They hold the practitioner. That condition does not change as AI capability increases. It becomes more consequential.

Large organisations are building AI workflows entirely on top of a single provider's stack. No internal capability. No IP. The dependency shifted from cloud hyperscaler to AI provider — same structural problem, new landlord.

The belief that keeps this in place: internal AI capability belongs to product companies. It does not. The organisations that own the judgment layer will not be exposed when the terms change. The ones who rented it will.

A consulting team that has spent twelve months producing AI-assisted deliverables is faster. They have become better at prompting. The question worth asking: have they become better at the underlying work — the diagnosis, the framing, the judgment?

If the answer is no, the leverage was borrowed. The debt is the gap between apparent capability and actual capability. It comes due when the tool changes.

The default in enterprise and consulting is Buy. COTS products, platform subscriptions, vendor solutions. Rational for commodity inputs. A structural problem when applied to capabilities that determine competitive position.

The signal that the wrong decision was made: a prototype converts to a product, and the conversation immediately turns to the team of three needed to maintain and run it. That is not a product. That is a support contract in human form.