When thinking gets cheap

A few years ago, a mid-sized trading company that wanted a decent market analysis of Southeast Asia had roughly three options: hire a consultancy, starting at six figures; recruit an analyst at HK$30–40k a month and wait a quarter; or have the boss pull a few all-nighters. Today a first draft of the same report takes an afternoon and costs close to nothing. It may not match the best human analyst — but it already beats most of the half-hearted work sold on the market.

The interesting part is not the speed. It's the price.

The third great cheapening

Over the past fifty years, business has been remade twice by something getting cheap. Computing got cheap, and the software industry appeared — along the way, every company became a technology company to some degree. Shipping and communication got cheap, and global supply chains appeared; "the world's factory" became possible. Both times, the firms that read the price signal early collected outsized returns. Most of the rest never quite saw what had happened.

This time it is thinking's turn — more precisely, the part of thinking that follows rules and patterns: drafts, translation, summaries, classification, basic analysis, routine code. The marginal cost of that work is falling from "the monthly salary of a trained person" towards "a unit of electricity".

Three ledgers get rewritten

For a business, this collapse in price rewrites at least three ledgers.

The first is cost structure. Knowledge work has always been a variable cost: one more client means one more stack of paperwork, one more trail of follow-ups, and eventually one more hire. When that layer goes to AI, the cost gap between serving ten clients and serving thirty flattens dramatically. For the first time, a small firm can serve clients to a big firm's standard — rare good news for SMEs.

The second is organisational shape. When one person plus AI covers what three people used to produce, "scale" changes meaning. What counts is no longer headcount but the density of judgement: whether each critical step has someone who genuinely knows the trade standing behind it. We expect the next decade to produce a wave of small, deep firms — few people, startling output per head. That model used to exist only in law firms and boutique investment houses. It is about to appear everywhere.

The third ledger is the easiest to miss: the moat has moved. When competence gets cheap, competence stops being a moat. The defensible ground retreats to two places — knowing what is worth doing, which is judgement about customers, markets and timing; and trust that others cannot supply: brand, relationships, the willingness to own the consequences when things go wrong. Neither can be generated from a model.

Output is getting cheap; judgement is getting expensive. When drafts are infinite, the person who can tell good from bad becomes the scarce one.

The price of standing still

A bucket of cold water before the close: none of this happens on the day a company buys its first AI subscription. Between the price signal and organisational capability sit process, incentives and habit — and that road has no shortcut. We wrote about the right order of operations separately.

But the direction is clear. AI matters to business not because it is an impressive tool, but because it has changed how the score is kept. Once the rules change, standing still is also a choice — it's just that the price of that choice goes up every month.