Sender rules: automated mail, forms and transcripts

Aide now sorts every incoming email and helpdesk message by who sent it before anyone works on it. Receipts, newsletters and other automated mail are filed out of the queue, contact forms are opened for the customer they carry, and your own rules decide the rest. Senders replaces the Spam and blocked settings. Every filed message says why in the inbox, and one click puts it right.

How it behaves

Rules first: Settings › Helpdesk › Senders lists your rules in the order Aide checks them, and the first match decides. A rule matches on the sender's address or domain, the recipients, the sender's name, the subject or the body (contains, starts with, or a pattern), the channel, the inbox, or who posted the message, with exceptions. Each rule shows how many messages it matched in the last 7 days.
Five outcomes: a rule files a message as from a customer, automated, ignored (Spam), a form submission, or a pasted transcript.
Built-in automated-mail detection: when no rule matches, Aide recognises mailing lists and newsletters, no-reply and notification senders, calendar invites, automated notices, and other tools' "we got your message" replies. You can switch it off on the Senders page. Everything else is a customer.
Automated mail: a new email conversation from an automated sender is closed straight away and found under the new Automated inbox view. Nobody is notified and Aide doesn't reply or draft. An automated email on an existing conversation is kept on the thread but doesn't reopen it or make it look like it's waiting on your team.
Ignored mail: goes to Spam as a new conversation, never onto an existing one, and nobody is notified.
Contact forms: for a sender you file as a form submission, Aide reads the customer's email and name from the labelled fields ("Email:", "Name:", and the same labels in other languages, or your own labels) and opens the conversation for that customer, so replies go to them and not to the form tool. Store data matched on that address is shown as unverified, so Aide doesn't act on their orders until they write from it themselves. A form with no customer in it stays with your team and Aide doesn't answer it.
Pasted transcripts: a chat a teammate pastes into a helpdesk conversation is read as the customer's words, so Aide can pick up the conversation from there.
On Zendesk, Front and Gorgias: the same rules and detection apply. Aide never answers an automated or ignored message, but never changes the helpdesk's status, tags or spam folder because of it; a conversation that starts with automated mail also shows under Automated. On a helpdesk conversation opened from a form, Aide drafts replies but doesn't send them, since the helpdesk would send them to the form tool.
Your existing spam lists carry over: blocked addresses and domains became a "Blocked senders" rule that ignores them, with your allowed domains as exceptions, exactly as before. If you only had allowed domains, they became an "Allowed domains" rule, switched off.
Test before you trust: "Test a message" classifies a pasted email or message and shows which rule decided and why. From a rule, you can replay it over your last 500 messages to see what it would change.
See why in the inbox: a message Aide filed carries a chip saying how ("Automated · Stripe receipts", "Mailing list or newsletter", "Form · via forms@wix.com", "Transcript", "Spam"), and the thread shows a line when a rule or the detection filed it. An automated conversation says which rule filed it above the thread, and the reply box says Aide won't answer (you still can). A customer taken from a form shows "From a form, email not verified" in the details panel.
Fix it in one click: the chip's menu has "This is a customer", "Not a customer (keep)" and "Ignore (Spam)", and the automated notice has "This is a customer". Bringing a filed message back reopens its conversation and runs what was skipped: topics, routing rules and the Copilot draft (not an AI reply or a notification, since you're already looking at it). Filing the only customer message away closes the conversation as automated or moves it to Spam. The thread records who corrected what.
"Always for this sender": admins can tick it when correcting. Aide adds the sender to one rule per outcome at the top of Senders ("Marked from the inbox: A customer", "…: Automated or internal", "…: Ignore"), and takes them out of the other two. They are normal rules you can edit.
Not spam: on a conversation a rule sent to Spam, "Not spam" corrects the message the same way, and offers "Always for this sender".

Also in this change

Newsletters and mailing-list mail are now closed under Automated instead of staying open in the inbox. With built-in detection switched off they stay in the inbox as before, and Aide still never answers them.
The "Treat bounces as spam" and "Aide marks spam" switches are gone; neither ever changed anything. Old links to the Spam settings page open Senders.
Senders has "Open Automated" and "Open Spam folder" shortcuts to the matching inbox views.

For developers

GET and PATCH /v1/spam-settings are removed. Sender rules are at /v1/sender-rules, the detector switch at /v1/sender-settings, and the tester at POST /v1/sender-classifications.
Conversations gain automated (list and detail) and requesterRelayed (detail).
Messages gain sender: how Aide filed an inbound message (outcome, decidedBy, ruleId, ruleName, reason, relayedBy), or null for other messages and ones from before this change.
GET /v1/conversations takes automated=true|false, and saved views accept an automated filter (is, boolean).
PUT /v1/conversations/:id/messages/:messageId/sender with { outcome: 'customer' | 'automated' | 'ignore', remember? } corrects an inbound message and answers with the message; remember is for admins. PATCH /v1/conversations/:id out of Spam does the same for a conversation a sender rule filed.
Conversation events of type sender_classified carry outcome, previous (set on a teammate's correction), messageId, and the rule's name or the detector's reason.

All changes