How to set up and optimize Fin for Zendesk
This guide will help you set up Fin AI Agent to enhance your customer support experience. By following these steps, you'll train Fin with your support content, optimize its responses, and deploy it in Zendesk to efficiently handle customer inquiries.
Executive summary:
The best way to think about an AI agent like Fin is to imagine you’re onboarding a new team member. Just like a new hire, Fin needs access to the right information, clear guidance on how to communicate, and the ability to learn from experience. By gradually providing Fin with relevant content, refining its behavior, and testing its responses, you ensure it effectively supports your customers. Fin will continuously improve, learning from its interactions and the adjustments you make to its training.
To deploy Fin as seamlessly as possible, we use a simple four-step framework: Train, Test, Deploy, and Analyze:
Train: Give Fin access to your support content, define its communication style, and set escalation rules to ensure it delivers accurate and helpful responses.
Test: Before launching Fin, validate its responses by asking real customer questions, refining content, and making necessary improvements.
Deploy: Set Fin live within Zendesk, whether in live chats, ticketing or a combination, and define how it interacts with customers and escalates issues.
Analyze: Use reporting tools to monitor Fin’s performance, optimize responses, and continuously improve how it handles customer inquiries.
This guide will walk you through each of these steps so you can set up Fin effectively and maximize its impact on your support operations.
Before you get started:
How Fin works in Zendesk
Fin acts as an AI-powered support agent within Zendesk, helping to resolve customer inquiries, automate workflows, and escalate issues when needed. It integrates directly with Zendesk’s knowledge base, messaging, and ticketing system to provide accurate, real-time responses. Depending on your setup, Fin can respond to live chats in Zendesk Messaging, handle customer tickets, or work through the Intercom Messenger while escalating to Zendesk agents when necessary. Understanding how Fin fits into your Zendesk environment will help you configure it effectively for your team’s needs.
Assemble your project team
To successfully deploy Fin, start by assembling a dedicated project team with key roles. The Support Leader ensures Fin aligns with customer support goals, while the Technical Leader manages integration and troubleshooting.
We also recommend having a content owner responsible for optimizing Fin’s knowledge base to ensure it has the right information.
Define success metrics
Once the team is in place, establish clear success metrics to measure Fin’s effectiveness. Focus on resolution rate and CSAT to assess answer quality and customer satisfaction, while average handling time and total hours saved offer insights into efficiency gains.
Additionally, consider qualitative factors like tone, clarity, and answer quality to ensure Fin aligns with your brand. While deflection rate and involvement rate can provide useful insights, they don’t always present a complete picture of Fin’s overall impact.
Step 1: TRAIN:
Train Fin with your support content
Fin provides accurate responses by learning from your existing support materials. Instead of answering from scratch, it pulls from your knowledge base to ensure customers receive consistent and informed responses. The better Fin’s training materials, the better it can assist customers. To get started:
Sync content from Zendesk: Connect your Zendesk help center to Fin so it can use your help articles for responses.
Upload PDFs or text snippets: Supplement your support articles with additional resources like FAQs or policy documents.
Manage content updates: Keep Fin’s knowledge base current by regularly syncing and updating content.
Train Fin with guidance
Beyond knowledge, Fin needs to understand how to communicate and when to escalate. By setting up guidance, you ensure Fin follows your company’s communication style and support workflows.
Set the basics: Choose Fin’s tone of voice and answer length, to match your support style.
Define communication style: Guide Fin on the vocabulary and terms it should use.
Set up clarification steps: Train Fin to ask follow-up questions when necessary to ensure the best possible response.
Establish escalation rules: Determine when and how Fin should escalate unresolved issues to human agents, ensuring seamless customer support.
Use tasks to automate multi-step processes
Fin can do more than just answer customer questions—it can also handle structured workflows, to automate repetitive or complex support tasks. Tasks allow Fin to complete multi-step processes while maintaining a natural, conversational experience.
Create task-based workflows: Automate troubleshooting steps, refund requests, or other structured processes.
Integrate with external systems: Enable Fin to pull in data or take action, such as updating order statuses or processing refunds.
Maintain a conversational experience: Ensure interactions feel smooth and intuitive, even when automating complex workflows.
[optimization step] Use content suggestions to improve your knowledge base
When you're ready to optimize Fin’s performance, this is a key step to consider. Refining your knowledge base ensures Fin can provide more accurate and comprehensive answers, reducing escalations and improving resolution rates.
As Fin interacts with customers, it will identify common questions that may not have sufficient coverage in your knowledge base. By leveraging content suggestions, you can continuously refine and expand your support documentation.
Identify content gaps: See which questions Fin struggles to answer and update your articles accordingly.
Create new support content: Get AI-driven recommendations for missing help topics based on real customer interactions.
Improve resolution rates: Ensure your knowledge base covers the most common customer inquiries, reducing escalations to human agents.
👉 How to optimize content with Fin [HCA]
Step 2: TEST
Test Fin’s answers
Before fully deploying Fin to customers, it’s crucial to test its responses and ensure they align with your support expectations. Testing helps identify gaps in Fin’s knowledge and fine-tune its behavior before it handles real customer inquiries.
Ask real customer questions: Use Fin’s testing suite to evaluate accuracy and response quality.
Review Fin’s answers: Identify areas where additional training or content updates are needed.
Refine Fin’s configuration: Adjust guidance, content settings, and escalation rules based on testing insights.
For all scenarios, compile a set of control questions for your team to assess, checking whether Fin provided the best response. A good starting point is 50 of your most frequent customer queries.
How to optimize Fin’s answers:
Based on your testing results, you may want to make adjustments to improve the quality and accuracy of Fin’s answer.
If Fin’s answer has missed the mark, it’s typically due to one of these reasons:
Gaps in content
Lack of guidance
Missing data
Here’s how to fix each of these:
Update or add content so Fin can provide a better answer next time—this can include syncing new content sources, updating an existing article or creating a snippet.
Add or review the guidance you have set, ensuring it follows best practice.
Ensure Fin has access to the right data by syncing attributes from Zendesk or connecting via API with external systems.
You can return to these steps anytime throughout your trial, as you work to optimize Fin’s performance.
Step 3: DEPLOY
Deploy Fin on Zendesk
When deploying, we recommend starting with a subset of your customers. Just like any new support agent, Fin shouldn’t handle every conversation on day one. This approach helps you test, monitor and improve easily within the trial period. To narrow down your test, we recommend defining:
A customer segment, such as free users
A specific topic, such as a product area
Once you’re confident in Fin’s abilities, you can deploy it in different ways within Zendesk. Depending on your needs, Fin can respond to live chats, handle support tickets, or work within both systems.
Option 1: The Intercom Messenger
Fin responds to live chats in the Intercom Messenger.
If Fin cannot answer, it escalates to a Zendesk agent or creates a Zendesk ticket.
To set this up, you can:
Begin by setting up the integration if you haven’t already
Configure your workflow, including the handover from Fin to your support team
Define your audience or topics
Install the Intercom Messenger
Option 2: Zendesk Messaging
Fin responds to live chats directly in Zendesk’s messaging platform.
If Fin cannot answer, it escalates to a human agent or creates a Zendesk ticket.
To set this up, you can:
Begin by setting up the integration if you haven’t already
Configure your workflow, including the handover from Fin to your support team
Define your audience or topics
Option 3: Zendesk Tickets
Fin replies to customer tickets in Zendesk
If Fin cannot provide an answer, it escalates to an agent
To set this up you can:
Begin by setting up the integration if you haven’t already
Configure your workflow, including the handover from Fin to your support team
Define your audience or topics
Step 4: ANALYZE
Analyze Fin’s Performance
After deployment, tracking Fin’s performance ensures it continues to improve over time. Reviewing analytics will help you optimize responses, refine escalation triggers, and maximize automation efficiency.
For overall support performance, we recommend using Zendesk reporting to monitor broader customer support trends. If you want to analyze Fin-specific performance, refer to the Fin performance report and content gaps insights, which are available within the Fin platform. These insights help identify areas where Fin may need additional training or content improvements.
Monitor conversation success rates: Use reporting tools to assess how well Fin is handling inquiries.
Identify areas for improvement: Review Fin’s accuracy and refine its training as needed.
Optimize workflows: Adjust automation and escalation rules based on performance insights.
By following these steps, you'll ensure Fin is effectively trained, tested, and deployed in Zendesk to provide seamless, AI-powered support. Just like a new team member, Fin will continue to learn, adapt, and improve over time—helping your customers faster and more efficiently.