Intercom

AI agents and copilot to streamline support

eesel AI instantly plugs into your Intercom, learns from your past tickets, macros, help center and more, and seamlessly joins in like a new teammate.

Training your AI with Intercom

Help Center

For public help centers:

  1. Choose the bot, or create a new bot, to connect with your help center.

  2. Click the integrations menu on the sidebar.

  3. Click Intercom.

  4. Click "Setup integration".

  5. Enter you help center URL and click "Start crawling"

The bot will then crawl your help center and start adding links to your bot to use as knowledge within minutes.

Create a Intercom AI Agent

eesel can automate your frontline support as an AI agent inside Intercom. You can have it leave internal notes, automatically reply, and triage tickets.

Once your AI has been trained and customized, it's time to link your Intercom domain to your AI bot.

To set up eesel as an AI agent inside Intercom, see our AI Agent guide.

Create a Intercom AI Copilot

Check out our setup guide for using AI Copilot in your Intercom.

eesel AI can act as an AI copilot for your Intercom by using our browser extension.

The AI can automatically pick up on the content of a ticket, and generate a response based on the knowledge it has access to.

You can use the browser extension to select the bot you want to use for its answer, and then draft a response and chat with the bot to edit before you paste is as a reply.

Intercom Actions

Once you've trained the bot and integrated it with your Intercom, it's important to customize it and add actions so that you can get the most out of your AI and automate manual tasks.

The following are Intercom specific actions. You can use them in combination with other actions on different platforms as well.

Tag conversation

intercom_tag_conversation

  • Purpose: Add tags to an Intercom conversation.

  • Key Params: tag: str

  • When/How to Use: Similar to zendesk_tag_ticket. Use for categorization, reporting, or triggering workflows based on conversation topics. Note that this tool takes a single string tag. The conversation ID is automatically available.

  • Example Prompt Instruction:

    Upon successfully answering a user's question about feature usage, you must call intercom_tag_conversation. Set the tag parameter to 'feature-inquiry'. If the user was asking about a known issue, make a separate call to intercom_tag_conversation with the tag 'known-issue'.

Close conversation

intercom_close_conversation

  • Purpose: Close an Intercom thread after a short delay (~30 seconds).

  • Key Params: — (No explicit parameters)

  • When/How to Use: Similar to zendesk_close_ticket. Use when the conversation is resolved and no further action is needed. The approximately 30-second delay is built-in. The conversation ID is automatically available.

  • Example Prompt Instruction:

    If the user confirms their issue is resolved and expresses satisfaction, or if the conversation has been inactive for several minutes after you provided a solution, you must call intercom_close_conversation to close the thread after the standard delay.

Leave internal note

intercom_leave_internal_note

  • Purpose: Leaves an internal note on the ticket.

  • Key Params: note: str

  • When/How to Use: Use this action to leave internal notes on conversations in Intercom. Describe what you want the internal note to be - this could be a draft reply for your agents to pick up, a translation of the user message and so on.

  • Example Prompt Instruction:

    If you escalate a conversation to a human agent, or if you need to summarize the AI's actions for the next agent, call intercom_leave_internal_note. The note parameter should contain the summary. For urgent escalations that need specific attention, you can also include notify_emails=['[email protected]', '[email protected]'] to alert specific team members.

Assign conversation (escalation)

intercom_assign_conversation

  • Purpose: Assign an Intercom conversation to a specific team or admin.

  • Key Params: assignee_id: str

  • When/How to Use: Similar to zendesk_assign_ticket. Use this for routing conversations that require human follow-up or expertise from a specific team. Define how the AI determines the appropriate assignee_id. This parameter requires the Intercom ID for the admin or team, not an email address. The conversation ID is automatically available.

  • Example Prompt Instruction:

    If the user asks a question about their billing history or requires an action related to their subscription, you must inform them you are escalating their request to the billing team and call intercom_assign_conversation. Set the assignee_id parameter to the specific Intercom ID for your 'Billing Team' inbox or a specific billing specialist admin.

Get ticket fields

Intercom_get_ticket_fields

  • Purpose: Retrieves specified fields from an Intercom ticket.

  • Key Params: fields: str (comma-separated)

  • When/How to Use: Use this when you need specific information from the current ticket (e.g., its current status, assignee, or a custom field value) to inform your next action or response. You must specify the field names (standard or custom field IDs as strings) you need. The ticket ID is automatically available.

  • Example Prompt Instruction:

    Before escalating a ticket, you should check its current status and assignee. Call intercom_get_ticket_fields with the parameter fields='status,assignee_id'. Use the results to inform the user about the current state or to decide who to assign it to next.

    If a user's query depends on a custom field (ID 456), call intercom_get_ticket_fields with fields='456'.

Get conversation fields

Intercom_get_conversation_fields

  • Purpose: Retrieves specified fields from an Intercom ticket.

  • Key Params: fields: str (comma-separated)

  • When/How to Use: Use this when you need specific information from the current conversation (e.g., its current status, assignee, or a custom field value) to inform your next action or response. You must specify the field names (standard or custom field IDs as strings) you need. The conversation ID is automatically available.

  • Example Prompt Instruction:

    Before escalating a conversation, you should check its current status and assignee. Call intercom_get_conversation_fields with the parameter fields='status,assignee_id'. Use the results to inform the user about the current state or to decide who to assign it to next.

    If a user's query depends on a custom field (ID 456), call intercom_get_conversation_fields with fields='456'.

Un-assign conversation

Intercom_unassign_conversation

  • Purpose: Use this action to unassign conversations in Intercom.

  • Key Params:

  • When/How to Use: Simply describe when you want the conversation unassigned, and the AI will handle this. This is useful for removing an assignee from a conversation, making it available for others to pick up.

  • Example Prompt Instruction:

User lookup

Intercom_get_conversation_fields

  • Purpose: Allow your bot to look up a Intercom user using their email address.

  • Key Params:

  • When/How to Use:

  • Example Prompt Instruction:

Update conversation

Intercom_update_conversation

  • Purpose: Use this action to update a conversation in Intercom.

  • Key Params:

  • When/How to Use: Define the conditions and details you want to change (such as status, priority, or assignee), and the AI will automatically update the conversation accordingly.

  • Example Prompt Instruction:

    If the user indicates their issue is resolved and you have confirmed it, or if the workflow requires setting the ticket status to 'solved', you must call intercom_update_conversation. Set the payload parameter to a JSON string that updates the status, e.g., payload='{"status": "solved"}'.

    If the user reports a high-priority issue you are escalating, you might call intercom_update_ticket with payload='{"priority": "high"}'.

    To update a custom field (ID 123), call intercom_update_ticket with payload='{"custom_fields": [{"id": 123, "value": "new value"}]}'.

PreviousFreshdeskNextFront

Last updated