Lindy AI is a cutting-edge workflow automation tool for enhancing customer support operations. Its standout feature is the ability to seamlessly integrate with your existing knowledge base, enabling it to automatically draft, send, or save email responses to user inquiries. It is highly customizable, making it ideal for building a tailored support agent that aligns perfectly with your business needs. Lindy helps you save time, maintain consistency, and ensure a high-quality customer experience with its intuitive interface and powerful automation capabilities while reducing the manual workload for your support team.

In this tutorial, we’ll build a custom support agent using Lindy, a powerful workflow automation tool. Lindy seamlessly integrates with your knowledge base to automatically draft, send, or save email responses to user inquiries. 

By the end of this tutorial, you’ll be able to:

• Set up your trigger
• Add initial conditions
• Link a knowledge base
• Refine the workflow with additional conditions
• Draft and send the email

Let’s get right into it!

Step 1 - Set up your trigger

The first step is to create a workflow for the customer support assistant. Go to Lindy AI and sign into your account. 

Click ‘New Lindy’ in the left sidebar and select ‘Start from Scratch.’

The trigger module appears in the workflow editor. Click it and search/select ‘Email Received’ as your first trigger. This ensures that the workflow is triggered when you receive an email.

Now, click the ‘Connect’ button to connect your account with Lindy. 

After establishing the connection, we need to filter events to ensure that only the emails containing specific keywords can be processed in the workflow editor. Otherwise, it can get expensive, and you will incur unwanted costs. 

Turn on ‘Filter events on this trigger’ by clicking the checkbox. Choose ‘Body’ for the filter, then select ‘Contains’ in the subsequent drop-down box. Specify keywords that occur in the body of the email as filter triggers. The email containing the specified keywords will be filtered into Lindy for an automated reply. 

Click the ‘+ Condition’ button at the bottom of the event filter, select ‘AND,’ and specify another event filter. This time, specify a filter with a specific keyword in the email subject. Specify keywords for the event filter. 

A quick tip: You can specify more filters by clicking ‘+ Condition Group.’ This allows you to define and configure email filtering based on labels, reply-to, and other conditions. 

Step 2 - Add initial conditions

Next, add a filtering condition to narrow the workflow further. In our example, we want to filter emails addressed to the customer service department of a local broadband service. 

Click the ‘Condition’ step in the workflow editor. 

Specify the condition that starts with ‘Go down this path if…’ You can define conditions such as ‘Go down this path if the email asks a question about the ABC broadband, specifically regarding connectivity issues, billing issues, cancellation, refund, or login issues related to the broadband services. 

Now, what Lindy will do is scan the email for specific keywords that meet the conditions you specified here. If the condition is met, the workflow will continue. 

Save the workflow by clicking the ‘Save’ button at the top right corner of the screen. 

Step 3 - Link a knowledge base

The most important step is to connect Lindy to your knowledge base so it can fetch the response for the specific questions asked in the email. A knowledge base can be a Google Drive document, a Notion, or a text document. It can also be a link to your knowledge base on the website or any link that can be used to fetch answers to emails. 

Click the ‘+’ button and select ‘search knowledge base.’

Link your knowledge base sources such as a Google Drive link, text document, or a Notion database. For example, we will use web pages as links to the knowledge base. Select the appropriate source in the subsequent pop-up dialog. 

If you are linking webpages, click ‘Website’ and add links to multiple webpages. Click ‘Add Single Link’ when adding knowledge base links. You can also click ‘Crawl’ so it can add the web pages automatically. An example of knowledge base structure is:

• Policies
• Common FAQs
• Refund and discount procedures
• Login troubleshooting steps

Click ‘Save’ after adding all the links.

Next, choose the default model for parsing the knowledge base. Leave other settings at default. 

Step 4 - Refine the workflow with additional conditions

To ensure Lindy provides accurate responses, add a condition to check the knowledge base for relevant content before proceeding. For instance, the workflow should stop if no relevant content is found to avoid generating unhelpful responses. However, if relevant content is available, the workflow can continue seamlessly, such as drafting the email using the identified information. This ensures meaningful and reliable interactions.

Add a ‘Condition’ to the workflow.

Specify the condition in the condition prompt box. You can say, ‘Go down this path if you find a response to the email in the contents of the knowledge base.’ 

Do not forget to click ‘Save.’

Step 5 - Draft and send the email

The final action in this workflow is to draft an email. You can use a predefined template or a custom prompt tailored to the situation to ensure the email is professional, relevant, and straightforward. For example, the prompt could be:

Prompt:

Your job is to respond to the user's emails using a knowledge base of company, user, and product information. Be concise and clear, and avoid adding any information not found in the knowledge base.

The email draft should include placeholders for dynamic fields such as the recipient's name, the issue they’re facing, or the proposed solution. This makes it easier to personalize each email while maintaining efficiency.

To uphold professionalism, include your company’s email signature, which may feature your logo, contact information, and any relevant disclaimers.

You have two options for handling the email once it's drafted:

1. Automatically send the email to streamline the process and provide quick responses.
2. Save it as a draft for manual review before sending, ensuring greater control over the content and tone of the message.

These settings ensure that your emails are not only accurate but also consistent with your company’s standards and branding.

Click the ‘+’ button and search for ‘Draft Email.’ Go to the settings and specify the ‘To,’ Bc, or Bcc fields if you want the email to be forwarded to the concerned departments. Leave the subject on default. AI will draft the subject for you. In the body, click the AI prompt and use the prompt provided above. 

Lastly, add your signature to the email and click ‘Save.’

That’s it. Let’s test the workflow. We will send an email to the email address mentioned in the email received module, and it will draft a response for us. 

Lindy drafted a response and saved it in the ‘Drafts.’ You can also configure it to reply directly to the email ID.

That's it folks. You've successfully automated customer support emails and saved a ton of time and resources responding to them. We encourage you to experiment with the workflow and add conditions for each question type you receive. It will ensure that your customers get a response with detailed info, or the email can be escalated for advance resolution.