As a Service Agent, Fin’s goal is to resolve customer queries quickly and accurately while providing guidance and reassurance throughout their journey.

Fin achieves this through two essential support agent skills:

This guide walks you through how to train Fin as a Customer Agent on these skills so it can automate support at scale, while staying true to your brand voice and policies.

Fin can resolve common queries and immediately help customers find answers without waiting on a human agent.

With this skill, Fin acts as a Service Agent by using your FAQ content, how-to guides, best practices, and basic troubleshooting resources to deliver clear answers. Fin follows your policies, knows when to escalate, and can classify and assign conversations effectively.

First, go to Knowledge and make sure you’ve enabled Fin with content covering FAQs, how-to guides, best practices, and basic troubleshooting.

Next, use Fin Attributes to automatically classify every conversation, like issue type, sentiment, or urgency.

For example, to classify conversation topic:

Fin Attributes give you precise control over which conversations Fin handles and which ones are passed to your team. By classifying every conversation, Fin knows which ones to escalate and you can configure these settings under Fin AI Agent > Train > Escalation.

For example, if you always want “Pricing” queries to be handled by your team, add an Escalation Rule to escalate if the Fin Attribute selected for “Topic” is “Pricing”.

Then go to Fin AI Agent > Train > Guidance and create guidance to train Fin on your brand tone and policies.

For example, create “Communication style” guidance which tells Fin how to handle sensitive conversations: “If a customer sounds upset, confused, or mentions a bad experience, respond gently and with empathy. Acknowledge the emotion before offering help.”

Or create “Content and sources” guidance which tells Fin which source it should use for troubleshooting a particular issue: “If a customer says they're experiencing an issue with saving a project, always use [📄 Public article: Troubleshooting projects] to walk them through the exact steps to resolve this issue.”

Now it’s time to get Fin involved in conversations by deploying it in your workflows. You can decide when Fin responds and how it escalates conversations which are unresolved or require escalation.

Fin can handle complex customer issues by guiding them through technical troubleshooting or gathering the right context before escalation.

With this skill, Fin acts as a Service Agent by using connected customer data to personalize support and walk users through troubleshooting steps. When escalation is needed, Fin ensures all relevant details are collected and escalated to the right team, making handoffs faster and smoother.

First, go to Settings > Integrations > Data connectors to set up connectors that access data in external systems such as your Statuspage, Stripe, or Shopify Storefront. These data connectors can then be used in your Fin Procedure.

For example, you could connect Fin to your Statuspage to check for unresolved incidents when a customer reports an issue.

Next, use Fin Attributes to automatically classify every conversation, like issue type, sentiment, or urgency.

For example, to classify conversation urgency:

Now go to Fin AI Agent > Train > Procedures and click New procedure.

Give the procedure a title such as "Troubleshoot issue" and click Create from scratch to enter the editor.

In the When to use this procedure section at the top of the editor, describe exactly when Fin should (and should not) trigger this procedure. A strong trigger includes specific criteria and exclusions. For example: "Use this when a customer reports a bug, error message, or technical issue with the product. Do not use it if the customer is confused by an existing feature, is asking about a new feature, or is seeking product guidance unrelated to a malfunction."

Now write your instructions in the editor. Each instruction step uses plain natural language — just type what you want Fin to do. Use @ inside a step to add tools like data connectors, attribute updates, or handoffs. For example:

Once you’re happy with the simulation results, click Set live to let Fin get to work! 💪