Teaching the Brain through your AI

Reading is half of Connected AI. The other half is teaching: your AI saving what you tell it, so the Brain keeps learning while you work instead of only when you open the portal.

This needs a connection at the Collaborate level - chosen when a token is created, or approved when you connect claude.ai or ChatGPT. Read-only connections can't change anything.

What your AI can save

• Facts, rules and lessons. "We're dropping the entry tier and going mid-market from October" or "we never discount below 15%" becomes something the Brain applies to every report, brief and answer from then on.
• Moments. Launches, campaigns going live, wins, changes taking effect - with the date they happened, so your timeline stays honest even if you mention it a week later.
• Meeting notes. Paste in (or have your AI fetch) the text of a call or strategy session, and the workspace reads it the way it reads an uploaded document: pulling out the durable points with the quote that supports each one.
• Answers to the Brain's own questions. The Brain keeps a list of things it would like to know. Your AI can walk you through them and save what you say.
• Work to do. Something you agree in conversation - "rewrite the Easter landing page headline, before the 5th" - can go straight onto your Working On board with its priority and deadline, rather than living only in a chat you'll never scroll back to. Your AI can add work; it can never tick something off or delete it.

Your AI is instructed to save only what you actually stated - not its own guesses or suggestions.

How it decides what to save

Exactly as the portal does when you chat with the Brain:

• Stated plainly and confidently, it's saved - and appears in the portal with "You told us through your AI assistant" as its source.
• Hedged, or in tension with something already known, it's queued as a question for you to confirm. Until you do, it doesn't count.
• Meeting notes always queue. Everything pulled from a transcript waits for approval - conversation is messy, and a walked-back idea shouldn't become a fact.

Nothing your AI saves is permanent-until-noticed: your account team sees every addition in the workspace's change history and can undo any of it.

Reviewing what's waiting

Two places, both on your dashboard:

• "Waiting on your review" on the Overview tab lists documents and meeting notes with insights pending. Click Review to approve, edit or reject each one - with its supporting quote - and anything from a connection is marked added by your AI assistant.
• Open questions (also on Overview) carry the statements waiting to be confirmed. Answer them there, or let your AI ask you and save the answer.

Nothing waiting? Neither section appears - they only show when there's something to do.

Listen mode

Rather than you remembering to say "save this", your AI can notice for itself. Each Collaborate connection has a listening setting on the Connected AI card, changeable at any time:

• Listening off - your AI only saves when you ask it to.
• Listen: offer first (recommended) - when you state something durable about the business, your AI offers: "that sounds like a lasting change - want me to save it to your marketing brain?" Nothing is saved unless you say yes.
• Listen: capture freely - your AI saves as it goes and tells you it did, without asking each time. Everything still follows the rules above, so uncertain things still queue.

Changes take effect the next time that AI starts a session. Listening never gives your AI access to anything else - it only changes how eagerly it offers to save what you've already said.

Correcting and flagging

When something the workspace believes is wrong, your AI can fix it in the moment:

• Correcting replaces it with your version straight away - you're the authority on your own business. The original is kept, and your account team can see exactly what changed and put it back.
• Flagging is for "that's not right but I'm not sure what is". The item stays in use, marked as under review, and lands in your open questions for a decision later.

The same actions are available by hand in Exploring the Brain graph and on "What we know about you".

Costs

Saving a fact or a moment is free. Reading meeting notes uses your workspace's monthly AI allowance, as does asking the Brain questions - see Usage & AI allowances. If an allowance runs out, your AI is told plainly and everything else keeps working.

Staying comfortable with it

• Start with Read-only, move to Collaborate once the reading habit has proved itself.
• Keep offer first on until you trust what your AI chooses to save.
• Skim Recent activity on the Connected AI card now and then - every save is listed there.
• Anything you don't like can be corrected, removed or undone; and Disconnect or Revoke ends a connection instantly.

Back to What Connected AI is & how it works, or Connecting your AI tools.