Your first hour
Connect a mailbox, set autonomy to hold replies, put three pieces of knowledge in, and read the first drafts. That is the hour. Everything else — the phone, prospecting, the pipeline, the data grid — is better after you have seen what Connect writes when it is grounded in your own material rather than in nothing at all.
Do these four, in this order#
- Connect one mailbox on
#/mailboxes— the busiest one, not a quiet test address. - On
#/autonomy, set email toask_before_sendso every outbound reply waits for a person. - On
#/knowledge, add three things: what you sell, what you do not do, and how you want to be addressed. - Leave it for a while, then open
#/needs-youand read every draft before deciding anything.
The busiest mailbox is deliberate advice. A quiet address gives you two messages in an hour and no sense of whether the drafts are any good; the real one gives you a spread of enquiries, complaints and noise, which is what you actually need to judge.
The step people get wrong#
It is the third one. A workspace with no Knowledge produces replies that are grammatical, courteous and useless, because there is nothing specific to ground them in. Three short entries change the character of every draft, and they take a few minutes:
- What you sell
- The products or services, in the words your customers use, not your internal names.
- What you do not do
- The single highest-value entry. It is what stops a confident answer to a question about something you have never offered.
- Tone and address
- Whether replies are formal, who signs them, and anything you always say — or never say.
How to know the hour worked#
| Check | Where | What good looks like |
|---|---|---|
| Mail is arriving | #/inbox | Real conversations, with the right sender names resolved to people |
| The mailbox is healthy | #/mailboxes | Health reported, not merely 'connected' — the two are different claims |
| Nothing has been sent | #/needs-you | Drafts are queued for approval, because that is what ask_before_send means |
| The drafts cite your material | any queued draft | Specifics from your Knowledge, not generic courtesy |
| Decisions are recorded | #/autonomy-audit | Each draft shows what was decided and under which rule |
Leave these until tomorrow#
- The phone. It has its own setup and its own test, and doing it while you are still calibrating email splits your attention. Set up the phone when email is settled.
- Prospecting. Discovery is only as good as the brief you write, and a brief written before you have read a single draft tends to be vague.
- Autonomy loosening. Moving a channel to
autonomouson day one means the first thing you learn about the drafts is what a customer thought of one. - The data grid.
#/datais genuinely useful and genuinely a distraction on day one.
If something goes wrong during the hour and you want everything to stop while you look at it, use the runtime switch rather than unpicking settings. It halts Connect immediately and changes nothing, so turning it back on restores exactly the configuration you had.
Questions#
Should I connect a test mailbox instead of a real one?
A test mailbox proves the connection and teaches you nothing else, because nobody writes to it. If you cannot use a live address, forward a week of real mail into the one you connect — the value of the first hour is in reading drafts written against genuine enquiries.
How many drafts should I expect in the first hour?
That depends entirely on how much mail the mailbox receives; there is no target. What matters is that the first sync has caught up and that the drafts you do see are answering the message above them rather than a summary of it.
Nothing appeared at all. What is the first thing to check?
The mailbox's health on #/mailboxes, not the inbox. A mailbox can be connected — the credentials work — and still be failing to fetch, which is why health is reported separately from connection state.