# Everything a brand knows, in one place any AI can read.

A Brand Brain gathers what lives in inboxes, old proposals and people's heads into one organized set of linked notes. Your team can search it. Your AI tools can answer from it, and show where each answer came from.

We run one on our own work, with project notes going back to 2001. It holds client history, so there's no live demo. This page shows how it works instead, with made-up examples.

**Where knowledge lives today:** Emails, Proposals, Project notes, Briefs, Decisions, Past work. They flow into one Brand Brain (cleaned, linked, in an open format), which your team use (searches it, gets up to speed with it) and ai tools use (answer with your facts and your voice).

## Five steps from a pile to a map.

Same steps for our brain and for yours. The pieces come together one step at a time.

1. **Gather.** We pull from wherever your knowledge lives now: project tools, shared drives, proposals, briefs, the email threads you pick and past work.
2. **Clean and redact.** Passwords, logins, personal details and anything private come out. The raw files stay where they are. Only cleaned notes move on.
3. **Write the stories.** For each project or topic we write a short, plain summary: what happened, what we decided, what it cost, what we'd do again. A missing fact gets marked as a gap, never guessed.
4. **Link.** Each note links to the client type, the services, the people's roles, the decisions and the other projects it touches. The links are what turn a pile of notes into a map.
5. **Store it openly.** Plain text files with a few labeled lines at the top. Any AI tool can read them, and so can you.

### What "open format" means

The brain is a folder of ordinary text files. Each one starts with a few labeled lines (what kind of note it is, a title, a date, what it links to), so a machine knows what it's looking at. The rest is plain sentences.

There's no special database and no software to license. It follows the Open Knowledge Format, a new public standard for exactly this kind of folder. Any AI that can read text can use it, and if you ever leave us, the brain goes with you and still works.

One note, made up. The lines between the dashes are the labels.

```
---
type: Decision
title: Launch in two phases
date: 2019-04-02
project: Store rebuild, 2019
related: Shopify builds, Project lead
---
Catalog and checkout went live
first. Recipes and the blog came
six weeks later. The holiday
deadline didn't leave room for
both.
```

## Every note is linked to what it touches.

Here's a small brain with made-up nodes. Node types: Client type, Project, Service, Decision, Role.

- **Specialty food brand** (Client type)
- **Industrial maker** (Client type)
- **Nonprofit** (Client type)
- **Store rebuild, 2019** (Project)
- **Site redesign, 2016** (Project)
- **SEO program, 2024** (Project)
- **Donor site, 2021** (Project)
- **Shopify builds** (Service)
- **WordPress builds** (Service)
- **SEO** (Service)
- **Launch in two phases** (Decision)
- **Keep the blog on WordPress** (Decision)
- **Fixed fee, not hourly** (Decision)
- **Project lead** (Role)
- **Designer** (Role)
- **Developer** (Role)

Example questions and how the answer travels:

### Why did the 2019 store rebuild launch in two phases?

- **Search.** Finds the project note for the 2019 store rebuild.
- **Follow the links.** Follows its links to the decision, the client and the project lead. Path: Store rebuild, 2019 → Launch in two phases → Specialty food brand → Project lead.
- **Answer, with sources.** The holiday deadline didn't leave room for everything. Catalog and checkout went first. Recipes and the blog followed six weeks later. The project lead made the call with the client in April. Sources: Project · Store rebuild, 2019; Decision · Launch in two phases; Role · Project lead.

### Have we kept a blog on WordPress next to a new site before?

- **Search.** Finds the decision note "Keep the blog on WordPress."
- **Follow the links.** Follows it to the two projects that made that call and the service behind them. Path: Keep the blog on WordPress → Site redesign, 2016 → Donor site, 2021 → WordPress builds.
- **Answer, with sources.** Twice. The 2016 site redesign and the 2021 donor site both kept their blogs on WordPress, to keep the old posts and their search rankings. Sources: Decision · Keep the blog on WordPress; Project · Site redesign, 2016; Project · Donor site, 2021.

### How do we usually price SEO work?

- **Search.** Finds the SEO service and the projects linked to it.
- **Follow the links.** Follows the 2024 program to its pricing decision and the client it was for. Path: SEO → SEO program, 2024 → Fixed fee, not hourly → Specialty food brand.
- **Answer, with sources.** As a fixed monthly fee, not hourly. The 2024 program set the pattern, and the decision note says why: the client wanted a number they could budget. Sources: Service · SEO; Project · SEO program, 2024; Decision · Fixed fee, not hourly.

## Four questions it answers every week.

Made up, but typical. Every answer comes back with the notes it came from, so you can check it.

### Drafting a proposal

- **Question:** We're quoting a store rebuild for a specialty food brand. What did similar projects cost, and what ran over?
- **What comes back:** Three similar rebuilds since 2019 took 8 to 12 weeks. Two ran over on cleaning up product data, so newer proposals price that as its own line.
- **Sources:** Proposal · Store rebuild, 2019; Project · Store rebuild, 2023; Decision · Price data cleanup separately

### What did we decide?

- **Question:** What did we decide about free shipping last spring?
- **What comes back:** Free shipping over $75, starting in April. Average orders sat just under $75, so the bar nudges people up. The note says to revisit after the holidays.
- **Sources:** Decision · Free shipping threshold; Meeting notes · Spring planning

### Getting a new person up to speed

- **Question:** I'm new. What should I know before my first call with this client?
- **What comes back:** They've worked with us since 2014. They prefer email to calls, approve all copy in writing, and their busy season runs October to December. Read these three notes first.
- **Sources:** Client map; Project · Site redesign, 2016; Playbook · Client handoffs

### Giving an AI assistant your voice

- **Question:** Write a product description for the new holiday gift box.
- **What comes back:** The assistant pulls the brand's voice notes (short sentences, no hype words), the product facts and the words to avoid, then drafts. A person reviews it before it goes anywhere.
- **Sources:** Voice guide; Product sheet · Holiday gift box; Words we don't use

## A brain is a loop, not a project.

A brain that's six months out of date is worse than none, because people stop trusting it. So the updates run in a loop.

1. **New work lands.** A project wraps, a decision gets made, a proposal goes out.
2. **We write it up.** We clean it, summarize it and link it, the same way as everything already in there.
3. **The index rebuilds.** Every morning the search picks up whatever changed.
4. **Freshness check.** We compare the brain with the project tracker and list anything new that isn't written up yet. That list starts the next loop.

Then back to step 1.

## What never goes in, and who gets to ask.

- **What never goes in.** Passwords, logins, keys, personal contact details and payment info. Raw email dumps stay out too. The original files stay where they are, and only cleaned notes go in.
- **Redaction, twice.** We scrub every note when we write it and mark it once it's checked. The search then masks anything that looks like a password and sets it aside for a person to review. A separate pass looks for people's names, because a password scan misses those.
- **Who can ask.** Only your team and the tools you approve, over a private connection. The search can read but never write: it can't change a note or send anything. There's no AI inside the search itself, so a cleverly worded document can't talk it into anything.
- **Answers show their work.** Every answer lists the notes it came from. Anything a customer or client will read gets checked by a person before it goes out.

## Our own brain, by the numbers.

Real counts from the brain we run on our own work. No client details, just its shape.

**From archive to brain:** About 25,000 files in our archive → About 3,600 text documents worth reading → 54 cleaned, linked documents in the brain → 78 links between them. We distill, we don't dump. Most of the archive is design files and old copies. The brain keeps what's worth knowing.

- **2001** oldest records it draws on
- **98** projects in our tracker since 2005
- **33** project stories written up
- **11** kinds of documents
- **5** sources searched with one question: project history, decisions, working notes, how-to playbooks and hand-off notes

**What's in the brain, by kind of document:** Project stories 33, Case studies 6, Website copy 4, Indexes 3, Other 8. 54 documents. Other includes the pricing reference, a playbook, the client and capability maps, an analysis, a reference note, the catalog and an overview.

## Your brand's brain, built in 4 to 8 weeks.

We build a Brand Brain for your company as part of our AI work: your voice, products, policies and past work in one open-format source that every tool and teammate can draw on. It's $5,000 and takes 4 to 8 weeks. You own it, and it works with whatever AI you use next. [Talk to us](https://nr10.com/contact/) · [See brand knowledge bases](https://nr10.com/what-we-do/ai-agents-and-automation/)

Human version: https://nr10.com/labs/brand-brain/