In short: There is no single right enterprise brain. There is a right one for two people, a different one for fifty, and a very different one for 10,000. The mistake at every size is building the brain of the wrong stage: too much too early, or too little too late.
Key points
- At two people the problem is false shared memory; at ten it is the first stranger.
- At fifty the problem is meaning; at two hundred it is delegation.
- At one thousand the problem is governance; at ten thousand it is density and federation.
- Read the problem, not the headcount: a regulated fintech may hit governance at 80 people.
- Revisit the diagnosis every time headcount roughly doubles.
Most writing about enterprise brains assumes a large, regulated organisation with platform teams, data stewards and a compliance function. Most organisations are not that, and the ones that are did not start that way. Every 10,000-person firm was once two people in a room, deciding things and remembering why, until one day they did not.
What changes as an organisation grows is not whether it needs a judgment record. It is which problem the record has to solve.
Six rungs, six problems
Two people. The problem is false shared memory. Two founders remember the same decision differently, and neither is lying.
Ten people. The problem is the first stranger. New hires do not share the founders' history, and every "why do we…" question routes to a founder.
Fifty people. The problem is meaning. Teams form, and the same words start to mean different things in different rooms.
Two hundred people. The problem is delegation. Founders no longer make most decisions; managers do, and the reasoning has to travel without them.
One thousand people. The problem is governance. Auditors, regulators, risk functions and security teams arrive, and "we decided it in a meeting" stops being an acceptable record.
Ten thousand people. The problem is density. There is finally enough volume in enough decision classes for a system to learn, and the challenge is federation: one organisation, many records.
The rungs are approximate. A regulated fintech may hit the governance rung at 80 people. A research lab may stay at the meaning rung for years. Read the problem, not the headcount.
Five things that change with size
- Decision volume per class. Low and heterogeneous when small; high and repetitive when large.
- Distance between decider and consequence. Zero at two people; several layers at 10,000.
- Turnover of context. Rare at first; constant later, through hiring, rotation and departures.
- External scrutiny. Investors at first; then customers, auditors and regulators.
- Who maintains the record. Founders; then everyone; then a function; then a platform.
Build for the next rung, not two rungs up
The practical rule is to build the layer the next stage will break, not the one the stage after that. A ten-person company that builds a governed ontology is spending money on a problem it does not have, and the ontology will be wrong by the time it matters. A 1,000-person company still running decisions through a founders' doc is about to discover the problem it skipped.
In practice
- Identify which rung your organisation is on by its problem, not its headcount.
- Name the problem of the next rung, and build only what solves that.
- Revisit the diagnosis every time headcount roughly doubles.
What would prove this wrong. Organisations that build the full large-scale stack early, and gain from it at every stage without excess cost or rework. Then the ladder is a false economy.
Common questions
How should a company's knowledge system change as it grows?
The record has to solve a different problem at each stage: false shared memory, the first stranger, meaning, delegation, governance, then density and federation.
Should a small startup build a governed ontology?
No. A ten-person company that builds one is spending on a problem it does not have, and the ontology will be wrong by the time it matters.
What is the scale ladder for enterprise knowledge?
Six approximate rungs at two, ten, fifty, two hundred, one thousand and ten thousand people, each with its own problem for the decision record to solve.
How do I know which stage my organisation is at?
Identify it by its problem, not its headcount, then name the next rung's problem and build only what solves that.