LAYER ONE · PHASE 1 · REQUIRED FIRST

Identity.

Who you are and how you are allowed to behave. Three documents that make every AI answer sound like you — and decide like you.

Why identity comes first

AI is already answering questions about you — to your customers, your tenants, your partners, and to you. If you have not written down who you are, the model fills the gap with the average of the internet. That is why AI output feels generic: it is generic, because it was never told otherwise.

Identity is the layer everything else stands on. Knowledge classification needs to know whose doctrine outranks whose drafts. Governance needs to know whose judgment the rules encode. The Operator needs a voice to stay on and principles to decide by. Skip this layer and every layer above it wobbles.

The test: could a stranger read these three documents and then write, decide, and refuse the way you would? If yes, this layer is done. If no, keep sharpening.

The three documents

Layer 1 is exactly three files. Plain text or markdown, on your machine, in a folder called 01_Identity/. No tooling required. An afternoon of honest writing beats a month of prompt tricks.

Document 01: Identity Charter

01_Identity/Identity_Charter.md

A non-negotiable statement of who you are — and will never become. One clear paragraph, then two short lists. No marketing language. Describe the actual operator or organization, the way you would to a sharp new hire on day one.

# Identity Charter

## Who we are
[One clear paragraph. No marketing language.
Describe the actual operator or organization.]

## What we stand for
-
-
-

## What we will never become
-
-
-

This charter is non-negotiable. All output and
recommendations must remain consistent with it.

How to write it well: the “never become” list is the sharp edge. “We will never pressure a tenant into terms we would not accept ourselves” does more governing work than any mission statement. Write at least three, and make each one something you have actually refused.

Document 02: Brand Voice

01_Identity/Brand_Voice.md

Exact rules for how the system writes and speaks. Not adjectives — rules. “Professional but friendly” is useless to a machine. “Lead with the point, no hype language, no unnecessary softening” is enforceable.

# Brand Voice

## Tone
- Direct
- Calm
- Precise
- Confident without arrogance

## Style rules
- Lead with the point
- Prefer short to medium sentences
- No hype language
- No unnecessary softening

## Vocabulary
Prefer: [words and phrases you actually use]
Avoid:  [words that never sound like you]

## Examples
On-voice:  " "
Off-voice: " "

How to write it well: the examples carry most of the weight. Paste in two real emails you were proud of and one AI draft that made you cringe, and label them. A model learns more from one on-voice/off-voice pair than from ten adjectives.

Document 03: Decision Principles

01_Identity/Decision_Principles.md

The five to nine rules that govern judgment when the answer is not obvious. These are the tiebreakers the system reaches for when two good options conflict.

# Decision Principles

1. Clarity over cleverness
2. Truth over comfort
3. Long-term over short-term
4. Authority must be explicit
5. Reversibility matters
6. Context before action

Add or rewrite these so they match how
you actually decide.

How to write it well: do not copy the defaults. Recall three hard calls you made in the last year and ask what rule you were following. If a principle would not have changed a real decision, it is a poster, not a principle.

Putting it to work immediately

You do not need to wait for the Operator to feel this layer pay off. Start any AI session by loading the three files (paste them, attach them, or point a project at the folder) and add one line: “All output must be consistent with these documents. Flag anything that conflicts.” The difference is immediate — drafts arrive in your voice, and recommendations stop contradicting positions you have already taken.

You are done with Layer 1 when

All three files exist in 01_Identity/ and are plain text an AI can read
The Charter's “never become” list has at least three entries you have actually enforced
Brand Voice includes at least one real on-voice and one off-voice example
Your principles would have changed at least one real decision this year
A draft written by AI with these files loaded sounds like you on the first pass

Next: the system knows who you are. Now it needs to know what it is allowed to trust. That is Knowledge — the vault and the authority levels that stop a rumor and a contract from being treated as equal.

← Foundations Layer 2: Knowledge →