Company BrainAI
Your company, in one place it can write from.
Fifteen modules, filled by reading your website, your documents and your own writing. Everything traced back to where it came from, and nothing used until you say yes.
It learns your company. Then it writes like you.
It reads your website before it asks you anything
You give it an address. It fetches the sitemap, samples the pages that matter, and pulls out what you already say about yourself — service lines, cases, regions, the words you use for things. Every extraction arrives marked as proposed, attached to the URL it came from.
- Public pages and documents only — nothing behind a login
- Uploads are extracted into anonymised records, then the file is deleted
- Every record carries its source and a confidence, never a bare assertion
You decide what is true, once
Records land in a table you can search, filter and rule on in bulk. Confirm what is right, edit what is close, reject what is wrong. Where two of your own pages disagree — and they will — it says so and asks you to pick, rather than quietly averaging them.
- Bulk confirm and reject, or edit the wording and keep the source attached
- Conflicts surface as a decision, with both sources and where each was seen
- Whatever you pick is applied everywhere that fact appears
It learns how you actually sound.
Connect a mailbox or upload a few things you wrote. Your website says how you want to sound; your sent mail says how you do. It reads sentence length, directness and the phrases that keep coming back, and writes them into a tone profile you can adjust.
- Read once, in memory — the messages themselves are never stored
- Words you use and words you never use, as rules the generator obeys
- Mark any passage as sounding like you, and the profile updates
What you type joins what it found.
Add a service line, a case or a number by hand and it does not sit in a separate box. It is merged with what the scan already knows, checked against it, and surfaced next time the generator needs that part of your company.
- Manual entries carry the same source trail as scanned ones
- Duplicates are matched and merged rather than stacked
- Anything that contradicts an existing record raises a decision
It learns who you sell to.
Your audiences, industries, regions and what makes you different are their own modules. The generator uses them to decide what a page has to answer and which of your differences are worth leading with.
- Audiences and industries, read from your own pages and cases
- What sets you apart, stated once and reused everywhere
- Regions and languages, so the writing lands in the right market
It gets sharper every month.
What people actually read, download and convert on comes back into the Brain. Subjects that earn attention move up; the ones that do not stop taking your time.
- Page performance decides which subjects deserve more
- Questions your leads asked become questions the next page answers
- Nothing here is guessed — if it cannot be measured it says so
A confirmed record is what makes a page publishable
When AnswerLink drafts a page, every factual sentence links to the record behind it. A sentence with nothing behind it does not produce a warning — it blocks the publish. That single rule is why the pages this produces survive being read by someone who knows the subject.
- Sentence-level citations, visible to your reviewer in seconds
- Unsourced sentences block publishing rather than nagging
- Confirming a record changes what every other module can say
What the Brain powers
One confirmed record changes what every other surface is allowed to say.
Every factual sentence carries the record behind it, one click from the draft.
An unsourced sentence blocks publishing. That is the whole reason these pages survive scrutiny.
The canvas proposes only territory your confirmed knowledge can actually defend.
See the theme canvasThe workshop skips everything the Brain already knows and asks only for what it cannot infer.
How it actually goes
What people ask about this
What if the scan gets something wrong?
It will. That is why nothing arrives confirmed. Extraction is a first draft of your company — the whole design assumes a person corrects it, which is faster than filling in an empty form and far more accurate than trusting a model's impression of your industry.
Do you keep the documents I upload?
No. A document is read, its insight extracted into anonymised records with names, amounts and counterparties stripped, and then the file is deleted. The source list keeps metadata — filename, type, when it ran, how many records it produced — and nothing else.
Can I add knowledge that is not on my website?
Yes, and most of the valuable material is not on your website. Add records by hand, upload documents, or paste in the LinkedIn profiles of the people whose expertise you are actually selling.
What happens when my company changes?
Re-read a source and it only adds what changed. Records that no longer match get flagged rather than silently overwritten, because a stale fact on a published page is worse than an obvious gap.
See your own Company Brain [AI] before you decide anything.
The scan is free and takes about four minutes. What comes back is your company, structured — not a demo of someone else’s.
Scan my website