---
type: "Article"
title: "Running a business on a second brain: what an LLM wiki replaces, and what it will not"
description: "Beyond notes: what changes when the same governed knowledge base holds your processes, your client context, your handovers and your decisions, and an honest account of where it stops."
resource: "https://www.laurelinlabs.com/blog/second-brain-business-operations"
tags: ["second-brain-for-business-operations", "aeo-guardrails-and-agentic-search"]
generated: { by: "laurelin-site-build/2026-09-02", at: "2026-09-02T12:00:00+00:00" }
status: "stable"
stale_after: "2027-03-02T00:00:00+00:00"
---
# Running a business on a second brain: what an LLM wiki replaces, and what it will not

Published 2026-08-27. Author: [Piers Butler](/entities/piers-butler.md).

A second brain becomes an operating model the moment the knowledge base stops being a place you file things and starts being the place work is defined. One governed set of Markdown files can hold your processes, client context, decisions and their reasons, and brand rules, and any capable AI agent can read all of it as working context. That gives an owner visibility without asking anyone for a status update, consistency without policing it, and handovers that survive a person leaving. It does not replace your accounting system, your CRM or your judgement, and pretending otherwise is how these projects fail.

## FAQ

### What size of business is this actually for?

It earns its keep fastest between roughly one and thirty people, where the knowledge is real but nobody has been hired to curate it. Below that it is still useful and mostly personal. Above it you are into governed systems with permissions and audit requirements that a folder of Markdown files was not designed to carry, though the layer still works well for a single team inside a larger organisation.

### Does this replace our project management or HR tools?

Not for anything with a legal or financial obligation attached. It replaces the layer of context those tools never held: why a decision was taken, what a process assumes, what a client agreed in a call eight months ago, who owns a handover. In practice it displaces the light-weight tools bought to hold documents and notes, and complements the systems of record.

### Can it hold financial or HR data?

Technically yes, since it is a folder on your own machine with no server behind it, which is exactly why the question is worth answering deliberately rather than by drift. Put in what improves decisions, such as run rate, utilisation, holiday capacity. Keep systems of record in the systems built for them, and never sync a vault holding personal data to anywhere you have not thought properly about.

### What stops the wiki becoming wrong over time?

A scheduled sweep and a shelf life. Claims about machine-checkable things (versions, paths, live URLs) are re-checked automatically. Positional claims like a stance or a style rule cannot be checked by a machine, so they carry a date and a review cadence instead. The important design point is that a value must be able to say never measured rather than defaulting quietly to zero, because a silent default is how a knowledge base starts lying.

### How do we get a team to actually use it?

Do not ask them to write it. That is the mistake every failed internal wiki makes. Their job is to feed sources in, transcripts, threads, documents, and to accept or reject the pages the agent drafts back. Start with one process that is currently person-bound and painful, prove the handover works, and let the rest follow.

Hub: [Guardrails, Agentic & Future-proofing](/topics/guardrails-future.md)