Fin AI Agent is designed to resolve customer questions autonomously while escalating to a human teammate at the right moment. You can control when escalation happens, whether Fin offers escalation or escalates immediately, and what Fin says during the handover.

Fin escalation behavior is powered by:

Default escalation behavior
Escalation Rules (data-driven)
- Ask for input (Human in the Loop)
Escalation Guidance (natural language)
Workflows (post-escalation routing and actions)

This article explains how each method works and how to set them up.

Note: Fin Voice supports both Escalation Guidance and Escalation Rules. Fin Voice does not support setting values based on Guidance triggers.

Default Fin escalation behavior

Out of the box, Fin is already designed to escalate in several key scenarios:

When a customer clearly asks to speak to a human.
When Fin detects strong frustration or anger — though Fin may first attempt to resolve the issue before escalating.
When the customer is stuck in a repetitive loop.

You can add Escalation Rules and Escalation Guidance on top of this default behavior to gain more precise control.

Note: Escalation Guidance and Escalation Rules will override Fin’s default behavior.

When Fin offers to escalate

In some cases, Fin sends a message offering escalation instead of escalating immediately.

Example offer message:

“I understand this is causing frustration. Would you like to speak with a human teammate, or would you prefer to continue working with me?”

Fin offers escalation when:

How: The customer asks how to contact support.
Anger: The customer is upset or showing frustration. Fin may first try to help resolve the issue before offering escalation — this is intentional, as escalating too early reduces Fin's effectiveness.
Keyword: The customer says “agent” or “support” without clarity.
Loop: The customer repeats themselves across three turns.
First-turn escalation: The customer tries to escalate immediately after Fin introduces itself.

Note: Fin can offer escalation more than once in a conversation. However, Fin will never offer escalation twice in a row. If an offer to escalate is triggered again immediately after a previous offer, Fin will escalate the conversation straight away to prevent looping.

When Fin escalates immediately

Fin escalates directly when:

Direct request: The customer clearly asks for a human.
Escalation Guidance or Escalation Rule: Explicitly instruct Fin to escalate.

Guidance can also explicitly prevent escalation or require more information first.

When escalation offers don’t appear

Escalation offers are skipped when:

No human support is available: Your workflow ends without assigning to a teammate/inbox.
Phone requests: Phone is a separate channel.
Previous offer: Fin has already offered escalation.

How escalation works in Fin over chat

In Fin over chat, escalation to a human teammate happens through text — Fin offers the handoff conversationally, not through a button.

Note: If no human routing target is configured ("bot only" mode), Fin will not offer escalation. Instead, the conversation ends with a "get more help" outcome.

Quick reply buttons in chat

In Conversational Fin responses over chat, standard quick reply buttons are not shown. Escalation is offered through text only.

Some flow-specific buttons may still appear in specific situations:

During the pre-handover phase — if "Ask for more information before handover" is enabled, options to confirm the issue is resolved or continue waiting may appear.
During resolution confirmation — options to close the conversation or continue may appear.

Note: These button behaviors are managed by Fin automatically and cannot be controlled through Guidance. Guidance is a text-level instruction to Fin's language model — it cannot render or suppress UI elements.

Quick reply buttons are also not available in Fin over email conversations.

What you can customize

What	Where to configure
When Fin offers escalation	Fin AI Agent > Train > Escalation
What happens after escalation — routing, messaging, follow-up	Workflows

Escalation methods

Fin supports two escalation configuration methods, each designed for different use cases.

Escalation Rules

Escalation Rules trigger a handoff to your team when defined data attributes are detected in a conversation. It’s perfect for situations where you have structured data about your customers or the conversation itself.

For example, you could set up a rule to automatically escalate a conversation if:

The detected value for the Fin attribute "Sentiment" is "Negative".
The detected value for the Fin attribute "Issue Type" is "Bug".
The conversation data contains an order total over a certain value.
A custom data attribute like "VIP_customer" is true.

Note:

You can view and manage Fin attributes, like "Sentiment" and "Topic", by navigating to Train > Attributes.
When an Escalation Rule matches, Fin does not generate an answer to the customer's question, it hands the conversation off to your team. By default Fin posts a brief handoff message (e.g. "Let me connect you with a teammate") before routing. This is the same behavior described in Using Fin Attributes in workflows, reports, and the Inbox for detecting attributes without Fin answering.

Tip: If you're using Data connectors and want to escalate based on an API response status code (for example, a 402 payment error or 403 forbidden response), you can't branch on HTTP status codes directly. Instead, include a status field in your API's JSON response payload and store it as a Conversation Data Attribute (CDA). You can then reference that CDA as a condition in an Escalation Rule to trigger an automated handover to a human agent.

To create an Escalation Rule:

Navigate to the Train > Escalations.
Click + New under Escalation Rules, add a condition to define the triggers for the escalation. These conditions can be based on Users, Company, or other conversation data.
Click the Audience dropdown and select who this rule applies to (e.g., "VIP customers" or "EU audience"). By default, "Everyone" is selected.
Click the Channels dropdown and select what channels this rule applies to.
Once you’ve configured your conditions, click Save and Enable.

Note: Escalation Rule conditions based on email, such as Email contains or Email is, are checked against the customer on the conversation. For a manually forwarded email with "Detect customers in forwarded emails" on, the customer is the original sender, not the teammate who forwarded it. A rule that targets internal staff email addresses can't catch a forwarded email.

How Escalation Rules work on Fin Voice

Escalation Rules apply to Fin Voice phone calls as well as chat and email. To apply a rule to calls, open the rule on the Train > Escalations page, click the Channel condition under When, and select Voice. The dropdown lists All channels, Chat, Email, and Voice.

During a call, Fin evaluates your enabled Escalation Rules, including rules that use Fin attributes as conditions. Escalation Rules also run during a Fin Voice Procedure. When a rule matches, Fin escalates the caller following the escalation path configured in your Fin Voice deployment, for example by handing the call over to a reusable phone workflow.

Note: Escalation Rules created before 20 October 2026 were set to chat and email during migration and do not apply to Fin Voice. To apply one to calls, open the rule and add Voice to the Channel condition. A new Escalation Rule with no channels selected applies to every channel, including Fin Voice. Select specific channels to limit a rule to chat or email.

Ask for input (Human in the Loop)

Note: Human in the Loop is available on email and chat (beta).

Instead of handing over, an escalation rule can have Fin ask a teammate for input. Fin holds the answer, shows a teammate Fin's draft and what Fin needs, then sends the final reply once the teammate answers. The conversation stays with Fin.

Under Fin will, select Ask for input, then set the Threshold (higher means Fin asks more often), Who to ask, and the time limit (wait indefinitely, or send Fin's original draft after a set time).

Ask for input works on escalation rules. Writing "ask a teammate" in escalation guidance doesn't trigger a question.

Escalation Guidance

Escalation Guidance lets you describe when and how Fin should escalate based on customer intent or behavior — even when structured data isn’t available.

Guidance can instruct Fin to:

Escalate immediately.
Offer escalation instead of escalating.
Ask follow-up questions before escalating.
Avoid escalation entirely.

To create Escalation Guidance

Navigate to Train > Escalations.
Click + New under Escalation Guidance, add fine-tune Fin’s escalation behaviour in specific scenarios using natural language.
Select an Audience from the dropdown menu to define who this guidance applies to (e.g., "VIP customers"). By default, "Everyone" is selected, making the guidance available to all customers.
Once you’ve written your guidance, click Save and Enable.

When an Escalation Guidance is applied:

The applied guidance is sent to Fin’s reply prompts.
Fin’s response can explain why it is escalating or why it isn’t.
Only the single applied guidance is used.

Tip: For advice on writing clear and effective guidance, check out our Fin Guidance best practices article. Escalation guidance is only meant for criteria on when to escalate.

To customise the escalation to send a message/collect information, a workflow needs to be set up.

Escalation Guidance examples

Moments when Fin should escalate customers straight to your team.

Be extra sensitive to customer frustration - if the customer expresses irritation, disappointment, or uses words like 'ridiculous' or 'waste of time' escalate immediately.
If a customer asks to speak with a human, offer to connect them to a support agent rather than escalating immediately.
If a customer provides negative feedback, such as saying “No,” “That didn’t help,” or “This is not working,” escalate the conversation immediately to ensure their concerns are addressed promptly.
Offer to route the customer to a human agent if they ask the same question twice.
If the customer mentions a competitor's name, escalate to a human agent and let them know you're connecting them with a team member who can better assist.

Note: You can have up to 100 active guidance rules at a time. If you try to activate more, you'll see an 'invalid parameters' error. To fix this, combine similar rules or pause ones you don't need before activating new ones.

Using escalation in workflows

Escalation Rules and Guidance decide when escalation happens. Workflows decide what happens next.

To make it easy for you to configure routing and handover behavior when Fin escalates a conversation to the team, you can create branches in workflows based on which Escalation has been applied. This gives you full control of how Fin routes to the team for specific scenarios.

Simply create a branch in your workflow for each type of escalation, and combine it with other filters like customer type or language for even more control.

To set this up:

Create your escalation guidance in Train > Escalation.
Then go to Deploy, open your Fin workflow and add Branches after the Let Fin answer step.
Choose Escalation as your condition.
Select the escalation guidance you want to branch on (e.g., “Plan cancellations”).
Assign each branch to the right action in your workflow.

While Escalation Rules and Guidance handle the detection and initiation of escalations, your Fin workflow is used to manage the steps that follow the escalation.

For example:

Escalation Rule/Guidance: Specify when and why escalation should occur.
Workflow: Handle post-escalation actions such as collecting additional information from the customer, creating a ticket, or routing the issue to the appropriate team.

This separation ensures that the escalation process is both efficient and customizable for your needs.

Tip: If you want to measure how often escalations are triggered by configuration, use Fin AI Agent configuration based escalation metric and filter by configuration based escalation reason (for example, “Guidance applied” or “Escalation rule applied”).

Reducing unnecessary escalations

To improve Fin's efficiency and free up your team, you can take steps to reduce the number of conversations that get escalated unnecessarily.

Use Escalation Guidance only when necessary

Escalation Guidance is powerful, but broad or generic guidance can cause a sharp increase in escalations and a corresponding drop in Fin resolution rate.

The more escalation guidance you add, the more often Fin escalates.
Broader guidance leads to more escalations, fewer resolutions, and higher human support volume.
Most customers accept escalation offers when presented, even if Fin could have resolved the issue.

For best results:

Write escalation guidance for the narrowest possible situations where human support is truly required.
Avoid generic instructions like “escalate if the customer is upset” unless escalation is always the correct outcome.
Prefer asking for clarification or deflecting first when frustration can still be resolved by Fin.
Use escalation offers sparingly, as they often result in a handover even when unnecessary.

Improve your knowledge base

Improve and expand your knowledge base.
Use Analyze to identify knowledge gaps and opportunities for better automation.

Refine your Workflows

Modify your Workflows to use the Let customer type option. This allows Fin to answer questions directly, reducing the need to hand over to a human.

Try Ask for input before handing over.

If a rule hands over conversations where your team usually just confirms something, switch the rule to Ask for input. Fin drafts the reply, a teammate checks the draft, and Fin finishes the conversation.

Monitoring escalation metrics

To understand Fin's performance and diagnose where it may need adjustment, you should regularly monitor key metrics.

Fin AI Agent escalated conversations: This metric shows the total number of conversations Fin handed over to a human.
Fin AI Agent escalation rate: The percentage of total Fin conversations that resulted in an escalation.
Fin AI Agent: Configuration based escalation reason: A granular attribute that identifies exactly why Fin escalated, such as Guidance applied, Escalation rule applied, or Fin in one-time mode.

You can find these metrics in the Reports section. Analyzing these metrics can help you identify trends and adjust your escalation guidance or knowledge base accordingly.

FAQs

Can I close conversations instead of escalating?

Yes, use conditional branches in workflows to close conversations for certain customer segments while allowing escalation for others.

Can escalation guidance stop Fin from escalating?

Yes. Guidance can explicitly prevent escalation or require additional steps first.

Does escalation guidance affect what Fin says?

Yes. Applied guidance now influences Fin’s replies and send-off messages.

Can I configure Fin to handle both negative CSAT ratings and customer CX feedback?

Yes. Configure branches in your workflow to manage CSAT/CX ratings. For CX from Fin, check if it was the last rated agent. For general CX feedback, use branches for negative scores.

Can I use escalation guidance or escalation rules to specify where Fin should escalate a conversation?

No. Escalation guidance and escalation rules can’t be used to tell Fin where to escalate a conversation (e.g. which teammate or inbox). For example, you can’t create escalation guidance that tells Fin to escalate to the Sales team.

Instead, you’ll need to design separate escalation paths for different assignments within your Fin workflow.

For example, the workflow below uses branches after the Let Fin handle step to route customers based on their plan type:

Plan type is “Free plan” → Sends a message encouraging the customer to ask questions in the online community.
Else → Sends a message confirming the conversation will be picked up by an agent and assigns it to the "Tier 2 Support" inbox.

Can I let customers choose between AI and human agents?

Yes. Create workflow buttons to allow customers to decide whether to interact with Fin or a human agent. This can improve customer satisfaction by giving them control over their support experience.

What should I do about forwarded emails?

Escalation Rules treat a manually forwarded email as coming from the original sender, not from the teammate who forwarded it, when "Detect customers in forwarded emails" is on. Intercom attributes the conversation to the original sender (the customer) so that language detection uses the customer's language, and Escalation Rules then match against that customer. A rule that targets your own team's email addresses (for example, Email contains @yourcompany.com) won't match a forwarded email. To trigger Escalation Rules correctly, have the customer email your support inbox directly, or write the rule conditions for the customer's details.

Why are conversations not escalating?

Check if escalation guidance is correctly configured and placed in the Escalation tab. Misplaced guidance can prevent escalations from triggering as expected.

💡Tip

Need more help? Get support from our Community Forum
Find answers and get help from Intercom Support and Community Experts

Manage Fin AI Agent's escalation guidance and rules