Automations

An automation is a small workflow: one trigger, then any chain of conditions, waits, and actions. Build it visually, test it on a real record, then switch it on.

Triggers

• Record events: deal created, updated, stage changed, won, lost, value or assignee changed; contact or company created or updated; tag added or removed; comment added.
• Time-based (checked daily): deal gone stale for N days, close date approaching, close date passed.
• Integrations: Telegram message received (with an optional cooldown so a busy chat fires once), campaign reply, campaign finished, meeting scheduled.

Attach filters to the trigger so it only fires when it should, for example “stage changed from Negotiation” or “value above 5,000”. Filters can use custom fields too.

On Telegram message triggers, the AI can classify the intent (buying signal, pricing question, complaint, churn risk…) and sentiment of the message, and you can filter on both. Classification uses AI credits.

Conditions, waits, and branching

• Condition nodes check the record’s live state, with AND/OR groups, and branch into true and false paths.
• Wait nodes pause the run for minutes, hours, or days, up to 90 days.

So “deal moved to Proposal → wait 3 days → still in Proposal? → nudge the assignee” is four nodes.

Actions

• Send a Telegram message: to a teammate through the Cavyro bot, or to the contact in their existing conversation.
• Send an email or an in-app notification to teammates, deal assignees, admins, or the contact.
• Send a webhook: generic JSON for your own backend or Zapier/Make/n8n, or preformatted for Slack and Discord.
• Change the record: move a deal to a stage, mark won or lost, update a field, add or remove a tag, add a comment.
• Create a deal from a scheduled meeting, linked to the meeting’s contacts.

Messages support {{variables}} like {{deal.title}}, {{contact.first_name}}, and {{deal.url}}, which links back into Cavyro.

Testing and publishing

• Test run: pick a real record and watch the full step log. Conditions evaluate for real, waits resolve instantly, and actions only preview what they would do. Nothing is sent or changed, and tests never count toward your quota.
• Saving is publishing: every edit creates a new version. Runs that are already in flight finish on the version they started with.
• Disabling is a kill switch: it also halts waiting runs, immediately.

Run history

Every run keeps a step-by-step log for 90 days: what fired, what each condition compared, what was sent, and why something failed or was skipped. Transient network errors retry automatically. When your monthly quota runs out, runs are recorded as skipped with the reason shown, so nothing disappears silently.

Limits and permissions

• Monthly runs: 500 on Starter, 2,000 on Pro, unlimited on Enterprise. Resets on the 1st. One-time run packs (500 to 10,000) top up the current month on Pro.
• You can create unlimited automations; only runs are metered.
• Everyone can view automations and history; admins create, edit, and enable them.
• Changes an automation makes are labeled as the automation in the activity feed, not as a person.