Automation
An automation rule is deterministic: an event happens, conditions are checked, actions run, immediately. There is no machine learning inside the engine; rules can call your AI provider as an action but do not depend on it. Owners and Admins create rules; Managers can view, enable and disable them and run dry runs.
Events
Section titled “Events”| Group | Events |
|---|---|
| Conversations | created, assigned, reopened, resolved, message received |
| Tickets | created, updated (status or priority changed) |
| CRM | contact created |
| Phone | call ended, call missed, voicemail received, callback requested, call topic confirmed, scoring completed, Manager overrode a score |
| Marketing | form submitted, booking created |
| Timers | N hours without a reply |
Conditions
Section titled “Conditions”AND / OR groups over attributes of the object: channel, status, priority, inbox, assignee, team, tags, subject, message text; the contact’s e-mail, tags, language, country and blocked flag; ticket status and priority; elapsed time; custom fields. Operators: equals, not equals, contains, not contains, starts with, is set, is not set, greater than, less than. A rule without conditions applies to every event of its type.
Actions
Section titled “Actions”| Group | Actions |
|---|---|
| Routing | assign an agent, assign a team |
| State | set status, set priority, add tag, remove tag, snooze, resolve |
| Communication | send a message from a template, add an internal note, send an e-mail, create a ticket |
| Integration | send a webhook (POST or GET, 10-second timeout, three retries at 1, 4 and 16 seconds, HMAC-SHA256 signed) |
| AI ✨ | suggest a reply, classify, summarise (shown only when an AI adapter is connected) |
Message and note templates accept {{contact.name}}, {{contact.email}}, {{conversation.id}}, {{conversation.channel}}, {{agent.name}} and {{account.name}}.
How rules run
Section titled “How rules run”- Every enabled rule whose event matches is selected and sorted by its position in the list.
- Each rule checks its conditions and runs its actions in order.
- A failing action is logged and the next action still runs. After repeated failures a rule is disabled automatically and Admins are told.
- An AI action without an adapter is skipped.
- Rules do not stop each other: every matching rule runs. If two rules assign different agents, the later one wins; a tag added by one and removed by another ends removed.
Loop prevention: actions performed by rules never trigger other rules; only customer messages and manual agent actions do.
Forward only: a rule applies to events after it is enabled. Existing conversations are not reprocessed.
Dry run
Section titled “Dry run”Run test evaluates a rule against a sample event or a recent real one and lists what would happen without doing it. Dry runs of AI actions show the estimated cost.
Templates
Section titled “Templates”Automation → Templates holds ready-made rules (route billing questions, escalate angry customers, close idle conversations, open a ticket for low-scored calls) that you copy and adjust.
Log and analytics
Section titled “Log and analytics”Every execution is logged with the event, the rule, each action’s outcome and timing. The analytics tab shows executions per day, failures and the most active rules.
Example: a ticket from a bad call
Section titled “Example: a ticket from a bad call”Event scoring completed; conditions topic = Complaint, score < 6, duration > 60 s; actions create ticket in the Support queue with high priority and add tag complaint. Phone calls never create tickets on their own; this rule is how you opt in.