Connect by JBRH Open Connect

First response to a new contact

Four things decide it, in order: whether the channel is on at all, what the autonomy rule at the narrowest applicable scope says, whether that person may be contacted, and whether the allowance is spent. A call is answered or refused in the moment; on email and messaging a reply can be written and held. Nothing counts as sent without the provider's acknowledgement.

Status
Available What this means
Audience
both
In the app
#/inbox, #/needs-you, #/autonomy
Last verified
Product version
6.3.2

What the other person experiences#

ChannelIf Connect may answerIf it may not
PhoneThe call is answered and the greeting is already speaking while the rest of the brief loadsA spoken refusal in the business's own name, then the call ends — the caller is never left listening to silence
EmailA reply is drafted and sent, with the provider's acknowledgement recordedA finished draft waits in Needs You. The sender sees nothing at all until somebody releases it
WhatsAppA reply goes out under the same rules as any channel, inside the provider's session windowThe draft waits, and the session window keeps running while it does
SMSInbound is handled, attached to the person, and STOP is honouredOutbound depends entirely on the configured provider, and the live carrier carries no SMS — the Phone screen says so rather than presenting a thread that cannot send

First contact, stage by stage#

  1. Trigger — a message, a call or a form arrives from somebody the workspace has no history with.
  2. External event — the provider delivers it: a mail push or poll, a messaging webhook, a carrier posting to the answer URL.
  3. Authentication and workspace resolution — the workspace is resolved and entered before any setting is read. Reading a line setting before this is how a switched-on line reads as off.
  4. Ingest — the provider's shape is normalised by the adapter that owns that provider, and nothing downstream ever sees the provider's own tables.
  5. Canonical record — a thread or a call, a message, and a contact. The record exists before any decision, so a refusal has somewhere to be written.
  6. Classification and reasoning — what this is, how urgent, and what it needs. Priority is not decoration: the engine reads it when it picks work.
  7. Knowledge, memory and rules — knowledge for grounding, memory at four tiers resolved narrowest-first, and standing instructions, all under fixed budgets so a call is not slowed by a long brief.
  8. Autonomy and approval — the four gates below. This is where an answer becomes a send, a held draft, or a refusal.
  9. Action through a provider — the one send boundary, shared by the engine and by a person, so both are subject to the same checks.
  10. Result — sent only with the provider's acknowledgement. Where that acknowledgement never arrives the state is uncertain, and it is shown as uncertain rather than guessed either way.
  11. Relationship, timeline and memory — a new person and, where there is one, a company; the exchange lands on the timeline whichever way the decision went.
  12. Audit, usage and Needs You — the decision is recorded including a refusal, the usage is metered, and anything needing a person is queued.

The four gates, in the order they are asked#

  1. Is the channel on? An off channel is not a slow channel. Nothing is drafted and nothing is queued, which is why a workspace that switched a channel off and forgot sees silence rather than a backlog.
  2. What does autonomy say at the narrowest scope? Rules resolve contact, then endpoint — one mailbox or one number — then channel, then workspace, and the narrowest applicable one wins. autonomous acts within the gates, ask_before_send queues every outbound for a human yes, draft_only writes and stops without even asking.
  3. May this person be contacted? Suppression, unsubscribe, complaint and do-not-contact are checked in one place before anything outbound. A memory row tagged as a block holds across channels and across future conversations, which is what makes a block mean something.
  4. Is there allowance left? When the daily allowance is spent, work is held rather than dropped and the read cursor deliberately does not advance. Advancing a cursor on a refusal is how a mail agent loses messages permanently.

What can go wrong on a first contact#

The reply is written but nobody sees it
draft_only does not queue anything for approval, by design. It exists so a workspace can watch what Connect would say without anybody having to keep saying no.
The greeting is slow on a call
The opening is spoken while the rest of the brief is still assembling, so a slow greeting points at the model session rather than at the amount of context.
A first-time enquiry sits behind older items
The queue is ranked rather than chronological. A first contact is a different kind of debt from a thread that has been running a week, and the ranking reflects that.
The person already exists
Identity resolution attaches a new address to an existing person rather than creating a second one. If two records did appear, merging is a human decision and preserves both sides.
Nothing arrives at all
Then the failure is before this flow: a disconnected mailbox, a quiet mailbox that authenticates and returns nothing, or a line that is not receiving.

Questions#

Can Connect answer a first-time caller and hold email replies at the same time?

Yes — autonomy is per channel, and can be set more narrowly still. A common arrangement is an autonomous phone line with email in ask-before-send, because the cost of a wrong sentence differs between a conversation and a written record.

Does a refusal reach the person who contacted us?

On a call, yes: a spoken message in the business's own name, then the call ends. On email and messaging, no message goes out — the refusal is recorded and surfaced to a person in the workspace instead, because sending an unsolicited we-cannot-help note to somebody who may be suppressed is worse than silence.

How is a first response different from any other reply?

Mechanically it is not — the same gates, the same send boundary, the same evidence rule. What differs is context: there is no relationship history to ground the answer in, so knowledge and standing instructions carry more weight, and the record created here is what every later exchange attaches to.