A company is an information-processing system
Most software treats an organisation as a set of records. It is more useful to treat it as something that perceives, remembers, decides, and acts, because that is what it actually does.
Ask what a company is and you will usually get an answer about people, or capital, or a legal structure. Those are true and not very useful. A more operational answer: a company is a system that takes in signal from the world, holds a model of itself, decides what to do next, and acts. Everything else, whether the org chart, the tooling or the quarterly rhythm, is scaffolding around those four functions.
Why the record-keeping model runs out
Enterprise software has spent forty years getting very good at one of those four functions: holding state. The ledger knows what was spent. The CRM knows what was promised. The HRIS knows who was hired. Each is an excellent record and a poor participant. None of them decide anything, and none of them know what the others know.
The workaround has always been people. A finance controller who remembers that this supplier was problematic in 2022. An operations lead who knows which approvals are theatre and which are real. This works, and it is enormously expensive, and it walks out of the building at five o'clock.
The four functions, made explicit
If you take the information-processing view seriously, the architecture writes itself. You need something that perceives, meaning it reads state continuously from every system rather than in nightly batches. You need something that remembers, not documents but decisions, with the reasoning attached. You need something that decides under an explicit policy rather than an implicit culture. And you need something that acts, because a decision that does not produce a write is just an opinion.
That is the shape of Cortexia, and it is not a coincidence. We did not set out to build four layers; we set out to ask what the minimum viable brain for an organisation would need, and four is what fell out.
What changes when memory is explicit
The most underrated of the four is memory. When precedent is written down in a structured form, meaning this decision, this evidence, this alternative rejected and this outcome, a category of organisational failure simply stops happening. You stop relitigating settled questions. You stop repeating mistakes that someone already made and learned from. New joiners inherit judgment rather than reconstructing it.
This is not knowledge management. Knowledge management produces wikis nobody reads. This is memory in the functional sense: retrievable at the moment of decision, by the thing making the decision.
The uncomfortable part
Making decisions explicit means writing down what your company actually optimises for, and organisations find this genuinely difficult. Not because the answers are secret, but because they are often unexamined. The first serious deployment conversation is rarely technical. It is someone in the room saying: wait, is that really our threshold?
That discomfort is the point. A company that cannot articulate its policy cannot delegate it to software or to people.