The Sourcing Machine

Ask a firm whether it makes or buys and you will get a strategy. Watch one, and you will get a mix. Every service line runs at some in-house fraction that moves slowly and expensively — redundancy one way, ramp-up the other — and the baseline depends on who you are. A small firm keeps most things in-house and minimises churn. A mid-sized firm has enough turnover that offloading the churn to a supplier pays, and buys in the new skills its own people are too busy running legacy tech to learn. An enterprise decides by spreadsheet. All three keep the core value work close.

The part the spreadsheet forgets is the make-threat. A credible in-house capability is what disciplines your suppliers; approach fully-outsourced and you lose it — no internal benchmark, a captured customer, and a unit cost that rises precisely because you can no longer walk away. Being all-bought is not the cheap corner it looks like.

Then the tide comes in. When AI collapses the cost of knowledge work, firms insource what they can and shed the rest — into expert pools that depress each domain’s wage. Some of those experts do not wait to be re-hired: they found AI-native firms in the field they came from, undercut the incumbents who trained them, and then compete each other toward the cost of compute. Every provider’s P&L is on the table — income, cost, unit price, margin — and when a margin cannot be held, the provider retires from the market.

The instrument runs all of it live: the mix, the price transmission, the squeeze, the tide. Every figure is an illustrative default, editable precisely because it is arguable. The cost of the work fell to the cost of the compute — and the livings went with it. Move the dials and watch the logic, and the livelihoods, shift.