Route closed chats to the right people using flexible rules. Each rule decides when to fire (checks), who to notify (recipients), and whether the chat stays in your inbox for follow-up.
How to get there: Go to Setup → Chatbot in the top menu → click your chatbot → Notifications tab.
Every time a chat closes, Boei evaluates all your notification rules against it. Rules are independent: a single chat can match zero, one, or all of them, and each matching rule sends its own notifications.
A chat that matches no rule is auto-closed silently. You can still find it under Inbox → Filters → Closed.
This replaces the old "Send to You" dropdown, which forced a single global choice. With rules you can, for example, email sales when a lead is captured, ping Slack when the AI struggles, and webhook every chat to your CRM, all in the same bot.
Click Add rule to start blank, or pick a template to prefill checks and recipients. Common templates:
Templates are starting points. You can add, remove, or edit any check or recipient after inserting.
Add one or more checks per rule. With two or more checks, choose:
Zero checks means the rule fires on every closed chat.
These are computed from the chat itself, no AI evaluation, no extra cost.
| Check | Fires when |
|---|---|
| Lead form was submitted | All required lead flow fields are filled in |
| Email or phone shared in chat | Visitor provided contact info during the conversation, even outside a lead form |
| AI escalated to a human | The bot handed off to a live agent |
| AI couldn't fully answer | AI health score is below 60 (see Health Score Reference) |
| Chat had at least N messages | Message count meets the threshold you set (default 10) |
| Chat lasted at least N minutes | Conversation duration meets the threshold you set (default 5) |
| Chat started from URL matching | Landing page URL matches your pattern (supports * wildcards, e.g. */pricing*, */checkout*) |
| Visitor is a returning customer | Same visitor identifier was seen in a prior conversation |
| Conversation language is | Detected language matches your dropdown choice |
Describe a condition in plain language and AI evaluates it against the transcript. Examples:
Each AI check costs a small amount of AI credit per closed chat, so use it when structural checks alone can't capture the intent. One AI check per rule is common; adding two is fine but doubles the credit cost per chat.
A rule can have any combination of the four recipient types, or none at all (useful when you only want to control inbox visibility).
| Recipient | What happens |
|---|---|
| Sends the transcript to any email address (yours, a colleague, a shared mailbox, a distribution list). Add multiple email recipients on the same rule if you want a group to see it. | |
| Chat agent | Sends the transcript to the agent's account email and auto-assigns the reopened thread to their inbox queue. Pick this over a plain email whenever a specific person owns the reply. The agent must be added under Chat Agents first. |
| Slack | Posts a short summary with a link to the full transcript to the Slack channel. Uses an Incoming Webhook, configured once per bot. |
| Webhook | POSTs the transcript JSON to a URL you choose. Standard for CRM sync, agency data warehouses, or wiring into Zapier/Make. See webhooks for the payload schema. |
The Chat agent recipient does two things that a plain email doesn't:
If multiple rules on the same chat pick different chat agents, the first agent to be assigned wins; later chat-agent recipients still get the transcript email but the assignment isn't overwritten.
Each rule has an Add to my inbox for follow-up checkbox:
Use "off" for archival or CRM-sync rules where you don't want the inbox to fill up with routine chats.
If a chat matches multiple rules, "on" wins: any single rule with the toggle on keeps the thread visible.
Per rule, expand Email content options to customize what email recipients see:
These options only affect email recipients. Slack and webhook payloads are unchanged.
Two rules:
Result: you get notified on real leads and on struggling chats, routine handled-well chats close silently.
Three rules:
*/pricing*" OR "AI check: visitor is evaluating a purchase" → chat agent = your senior sales rep, keep in inboxResult: sales and support each see the chats they care about, CRM gets everything, the inbox stays focused.
Two rules per bot:
#client-nameResult: the agency gets full data, the client sees only their leads.