# Business Brain — context pack template

Version 1.1 · last reviewed 9 November 2026 · part of [The Business Brain](/free/business-brain/)

This is the structure the Business Brain skill fills in. You can also fill it by hand. Either way, the rules are the same: one file, 15 sections, 900 to 1,800 words, and every content line tagged.

The skill reads this file before it writes the Brain (Step 4), so the worked example below follows every rule it enforces. Copy the example's habits, not its content.

## The three tags

| Tag | Meaning | What the AI does with it |
|---|---|---|
| `[FACT]` | You stated it and can stand behind it | Uses it as true |
| `[ASSUMPTION]` | A hypothesis, an inference or something not yet tested (an unpaid price is an assumption) | Uses it, but says when an answer depends on it |
| `[GAP]` | Unknown | Never fills it; asks you or writes `[NEEDS INPUT: ...]` |

### Where the tags go

- **Line tag:** the default. Each bullet starts with its own tag: `- [FACT] Not included: logo design, photography`.
- **Section tag:** when every line in a section has the same status and source, put the tag once in the section heading, with the source: `## 5. Voice [FACT — from the owner's own writing]`. It covers every line and table row in that section. Any line that differs carries its own tag, which overrides the heading.
- **Not content:** the title, the `Owner:` header line, the "How to use this file" line, headings and table header rows need no tag.

A content line with neither a line tag nor a section tag is a defect. The skill's self-check rejects it.

The tags are the most important design decision in this file. Without them, a model turns your guesses into confident claims, and that is how invented proof ends up in front of a buyer.

## Word budget per section

| Section | Target words | Never cut below |
|---|---|---|
| 1 Snapshot | 60 | 40 |
| 2 Offers | 150 to 300 | the main offer in full |
| 3 Ideal client + Not a fit | 150 to 250 | 3 verbatim client phrases and the not-a-fit list |
| 4 Proof | 30 to 120 | one line, even if it says "none yet" |
| 5 Voice | 250 to 400 | 3 do and 3 don't rules with examples |
| 6 Pricing and terms | 60 to 150 | payment and refund terms |
| 7 Client path | 60 to 120 | first-reply target |
| 8 FAQs and objections | 120 to 250 | top 3 |
| 9 SOP index | 40 to 120 | the list, even without links |
| 10 Decision rights | 60 to 150 | the full table |
| 11 Never say / never do | 60 to 150 | the full list |
| 12 Data rules | 40 to 100 | the full list |
| 13 Current priorities | 40 to 80 | the first AI job and its success check |
| 14 Open questions | as needed | every GAP |
| 15 Change log | as needed | every version |

When the file runs long, cut voice examples and FAQs first. Never cut boundaries, decision rights or data rules: those are the lines that stop expensive mistakes.

## The blank template

Copy everything in the block below into a file named `business-brain.md` and replace each bracketed instruction. Replace every `[TAG]` with `[FACT]`, `[ASSUMPTION]` or `[GAP]`. A section whose heading carries a tag (sections 5, 9, 10 here) needs no line tags, unless a line differs.

```text
# Business Brain — [business name] — v1.0 — [date]
Owner: [name] · Track: [LAUNCH or OPERATE] · Next review: [date + 30 days]
How to use this file: Treat [FACT] lines as true. Treat [ASSUMPTION] lines as working hypotheses and say when your answer depends on one. Never fill a [GAP] yourself: ask, or write [NEEDS INPUT: ...]. A tag in a section heading covers every line in that section.

## 1. Snapshot
- [FACT] We help [specific client] with [specific problem] so they get [concrete outcome].
- [FACT] Stage: [idea / first clients / established since year]. Team: [who]. Markets: [countries, languages].
- [FACT] Business model: [one-off projects / retainers / subscriptions / products].

## 2. Offers
### [Offer name]
- [TAG] For: [who]
- [TAG] Deliverables: [step by step]
- [TAG] Not included: [at least two exclusions]
- [TAG] Timeline: [duration]
- [TAG] Price: [amount or range] — [tested: paid by real clients / untested]
- [TAG] Client must provide: [inputs, access, decisions]

## 3. Ideal client
- [TAG] Situation: [role, size, what is going on]
- [TAG] Trigger moment: [the event that makes them act now]
- [TAG] Their words: "[verbatim phrase]" · "[verbatim phrase]" · "[verbatim phrase]"
- [TAG] Already tried: [alternative] — failed because [reason]
- [TAG] Buying criteria: [what they check before saying yes]
### Not a fit
- [TAG] [type of client] — warning sign: [signal]

## 4. Proof we can use
- [TAG] [item] — source: [where it lives] — permission: [yes / not needed / not yet]
(If none: "[FACT] No client proof yet. Do not imply any.")

## 5. Voice [FACT — from the owner's own writing]
- Scales (1 to 5): formal-casual [n] · reserved-bold [n] · simple-technical [n] · serious-playful [n]
- Close to: [person or brand] · Never like: [person or brand]
### Do
- [rule] — e.g. "[real line from the owner's writing]"
### Don't
- [rule] — not "[counter-example]" but "[fixed version]"
### Words
- We use: [list]
- We never use: [list]
- Format: [language and spelling] · emoji [yes/no] · typical length [n] · sign-off "[...]" · address clients by [first name / title]

## 6. Pricing and terms
- [TAG] [price line] — [tested / untested]
- [TAG] Payment terms: [deposit, schedule, method]
- [TAG] Discounts: [policy] · Refunds: [policy]

## 7. Client path
- [TAG] [lead source] → [first reply, target speed] → [call or no call] → [payment] → [onboarding] → [delivery] → [follow-up]

## 8. FAQs and objections
- [TAG] Q: "[client's words]" — A: [owner's answer]
- [TAG] Objection: "[client's words]" — Answer: [honest answer]

## 9. SOP index [TAG — source]
| Task | Owner | Frequency | SOP location |
|---|---|---|---|

## 10. Decision rights and human checkpoints [FACT — set by the owner]
| Decision | AI may | Human must | Who |
|---|---|---|---|

## 11. Never say / never do
- [FACT] Never invent results, testimonials, statistics or client names.
- [TAG] [claim or topic] — reason: [why]

## 12. Data rules
- [TAG] Never into AI tools: [categories]
- [TAG] Only after anonymising: [categories]
- [TAG] Approved AI tools: [names]

## 13. Current priorities
- [TAG] 90-day priority: [one]
- [TAG] First job for AI: [recurring job] — success check: [how you will know it works]

## 14. Open questions
- [GAP] [question] — matters because [reason] — I will find out by [method] before [date]

## 15. Change log
- v1.0 · [date] · created from interview
```

## Section guide, with a worked example

**Fictional example.** Sam Okafor is an invented freelance web designer who is launching a service building websites for independent physiotherapy clinics (LAUNCH track). Sam, the clinics and every number below are made up for illustration. Nothing here describes a real person or result.

Watch the tags. Sam has no clinic clients yet, so anything about how clinics behave is an `[ASSUMPTION]`, even when Sam feels sure. What Sam decided (deliverables, exclusions, rules) is a `[FACT]`, because Sam stated it and stands behind it.

### Section 1 — Snapshot

Good looks like: a stranger could repeat it back after one read. Common mistake: listing everything you could do instead of what you sell.

```text
## 1. Snapshot
- [FACT] I build websites for independent physiotherapy clinics (1 to 5 practitioners) so new patients can understand the treatments and book online without phoning.
- [FACT] Stage: first clients; I work alone. Market: UK, English only.
- [FACT] Business model: fixed-price website projects, plus an optional monthly care plan.
```

### Section 2 — Offers

Good looks like: deliverables a client could check off, plus exclusions, with untested numbers marked. Common mistake: no exclusions, so the AI promises everything.

```text
## 2. Offers
### Clinic Website Build
- [FACT] For: independent clinics with 1 to 5 practitioners.
- [FACT] Deliverables: 6-page site, online booking connected to the clinic's existing booking tool, treatment pages written from a 45-minute interview, basic on-page SEO, 2 rounds of revisions.
- [FACT] Not included: logo design, photography, paid ads, booking-software subscription fees.
- [ASSUMPTION] Timeline: 4 weeks from content interview — untested, no clinic project delivered yet.
- [ASSUMPTION] Price: GBP 2,400 fixed — untested, no client has paid this yet.
- [FACT] Client must provide: practitioner bios, treatment list, booking-tool admin access (shared through the tool's own team invite, never a password in chat).
```

### Section 3 — Ideal client and Not a fit

Good looks like: client language in quotes, a trigger event, and a real not-a-fit list. Common mistake: "small businesses who want to grow".

LAUNCH: no clinic clients yet, so these are hypotheses. They become `[FACT]` in an UPDATE run once real conversations confirm them.

```text
## 3. Ideal client
- [ASSUMPTION] Trigger moment: the receptionist leaves or cuts hours, and phone bookings start going unanswered.
- [ASSUMPTION — public source] Their words: "patients keep asking if we do sports massage", "our site looks like 2012", "I don't have time to write anything".
### Not a fit
- [ASSUMPTION] Clinic chains with an in-house marketing team — warning sign: asks for a procurement form before a call.
- [ASSUMPTION] Anyone who wants medical claims on the site that their professional body would not allow.
```

### Section 4 — Proof we can use

Good looks like: honest, sourced, permitted. "None yet" is a valid entry and protects you.

```text
## 4. Proof we can use
- [FACT] No client proof yet. Do not imply any.
- [FACT] Portfolio: two sites I built for friends' businesses (not clinics) — links in portfolio.md, permission given.
```

### Section 5 — Voice

Good looks like: rules with real examples you wrote, plus counter-examples. Common mistake: adjectives only ("friendly, professional"), which every model already defaults to. Here one section tag covers every line, because all of it comes from Sam's own writing.

```text
## 5. Voice [FACT — from Sam's own emails and posts]
- Scales: formal-casual 4 · reserved-bold 2 · simple-technical 2 · serious-playful 2
### Do
- Lead with the patient's experience. e.g. "Your patient is on their phone at 10pm with a sore knee. Can they book before they give up?"
- Short sentences, one idea each. e.g. "Your site has one job. Get the booking."
### Don't
- No marketing hype. Not "Transform your clinic's digital presence!" but "Patients can book without calling."
- No medical claims. Not "cures back pain" but "treatment for lower back pain".
### Words
- We use: patients, bookings, treatment pages
- We never use: synergy, cutting-edge, game-changer, "leverage"
- Format: UK English · no emoji · emails under 120 words · sign-off "Cheers, Sam"
```

### Sections 6 to 9 — Pricing, client path, FAQs, SOP index

Good looks like: the terms an assistant would need to answer an email without asking you. Common mistake: pricing scattered across sections with different numbers.

```text
## 6. Pricing and terms
- [ASSUMPTION] Payment: 50% to start, 50% before launch — untested.
- [GAP] Refund policy: not decided yet (see section 14).
## 7. Client path
- [FACT] Instagram or referral → reply within 1 working day → 20-minute call → proposal → deposit → content interview.
## 8. FAQs and objections
- [FACT] Q: "Can you also manage our Google Ads?" — A: No. I can recommend someone once the site is live.
## 9. SOP index [ASSUMPTION — expected tasks, no clients yet]
| Task | Owner | Frequency | SOP location |
|---|---|---|---|
| Content interview | Sam | Per project | not written |
```

### Section 10 — Decision rights and human checkpoints

Good looks like: the AI knows exactly where its job stops. This section matters most once the Brain powers an agent or automation.

```text
## 10. Decision rights and human checkpoints [FACT — set by Sam]
| Decision | AI may | Human must | Who |
|---|---|---|---|
| Reply to a new lead | Draft | Approve and send | Sam |
| Quote a price | Repeat the price in section 6 | Approve any change or discount | Sam |
| Publish a post | Draft | Approve and publish | Sam |
| Health-related wording on a client site | Flag it | Get the clinic's sign-off | Clinic owner |
```

### Sections 11 and 12 — Never say and data rules

Good looks like: specific, with a reason. These two sections are what make the Brain safe to hand to an assistant or agent.

```text
## 11. Never say / never do
- [FACT] Never invent results, testimonials, statistics or client names.
- [FACT] Never promise more patients, bookings or revenue. Reason: I cannot control it and have no data.
- [FACT] Never use patient names, photos or stories without written consent from the clinic.
## 12. Data rules
- [FACT] Never into AI tools: patient data of any kind, clinic booking exports, passwords.
- [FACT] Approved AI tools: Claude (paid plan), with chat history settings checked.
```

### Sections 13 to 15 — Priorities, open questions, change log

```text
## 13. Current priorities
- [FACT] 90-day priority: first 3 paying clinics.
- [FACT] First job for AI: draft replies to enquiries the same day — success check: I send 4 of 5 drafts with only light edits.
## 14. Open questions
- [GAP] Is GBP 2,400 the right price? — matters because every proposal uses it — I will find out from the first 10 conversations.
- [GAP] Refund policy — matters because the first proposal needs it — I will decide before sending proposal #1.
## 15. Change log
- v1.0 · 9 Nov 2026 · created from interview
```

## Track notes

**LAUNCH** (idea or first clients): many lines will be `[ASSUMPTION]`, and that is correct. The Brain is still useful because it stops the AI from inventing proof and keeps every draft consistent. Re-run the skill in UPDATE mode after your first 10 conversations.

**OPERATE** (running business): this file is your Company Context Pack. Add team roles to section 1, link real SOPs in section 9, name a person for every row in section 10, and list your actual data-protection obligations in section 12 (check them with a qualified adviser in your country; the template is not legal advice). Assign one owner and a monthly review date, or the pack goes stale with your next price change.

## Quality checklist

- [ ] Every content line has a tag: its own, or a section tag in the heading. No hypothesis is tagged FACT.
- [ ] Every number, price and claim traces back to something I said or supplied.
- [ ] Section 3 has at least three phrases in my clients' own words (or a GAP saying I need them).
- [ ] Every voice rule has an example I actually wrote.
- [ ] "Not a fit", decision rights, never-say and data rules are filled in.
- [ ] No passwords, keys, account numbers or client personal data anywhere in the file.
- [ ] Under 1,800 words.
- [ ] Owner and next review date are set.
- [ ] One copy only: every tool reads this file, and exports are regenerated from it.