How to Build a Customer Support Email Agent with Lindy AI
Prefer to watch? Watch on YouTube · Subscribe for new builds · Read the transcript
In this video I build an AI agent with Lindy AI that watches my Gmail inbox, answers common customer questions from a knowledge base, and sends me a Slack DM when it doesn't know the answer. It's for small business owners who are tired of typing the same replies over and over. I recorded this in August 2025, and tools change fast, so check Lindy's current docs before you copy my exact steps.
What I cover in this video
- 0:00 Why I need this: my wholesale clothing business is getting more emails than I can keep up with, and Lindy has a customer support email agent template.
- 0:46 The prompt I give Lindy to describe the agent I want.
- 3:05 Using a Claude project to write the knowledge base for our wholesale Lululemon program.
- 5:23 The flow Lindy builds: email received, search the knowledge base, reply or escalate to Slack.
- 6:55 Connecting Gmail and adding the knowledge base from Google Drive.
- 9:12 The first test email, which fails with an error.
- 10:46 Fixing the trigger filters and the reply step, then adding context in settings.
- 14:35 The working replies the agent sends back to my test emails.
Starting from Lindy's template
I run another business, and we've been doing a lot more clothing wholesale lately. That means a ton of emails with the same questions: how much is it per unit, what's in the bundle, is it leggings, tops, or bras. Most of the answers are already in the email I sent, but people ask anyway.
Lindy has a template called customer support email agent, so I start there. I like Lindy because it's plug and play, and I had no experience with AI tools a few months before this video. After picking the template, I describe what I want in plain English:
I want an agent that will monitor my support inbox. It will help answer customer questions using a knowledge base. If the agent does not know the answer, then it will just escalate to a Slack direct message.
Lindy asks three things: which email provider I use (Gmail), whether I already have knowledge base content, and to connect Slack for the escalation. Then it lays out the flow and asks me to confirm it:
- An email is received in Gmail.
- The agent searches the knowledge base for relevant information.
- If it finds an answer, it replies to the email.
- If it doesn't, it sends me a Slack DM.
The whole flow was built in less than two minutes. One note: I use GoHighLevel as my CRM and I couldn't find a Lindy integration for it, so for now I'm forwarding those emails to Gmail so they still get answered. If you want more on how agents like this fit together, I keep a page on AI agents.
Writing the knowledge base with Claude and a Google Doc
The agent is only as good as what you feed it. I already have a Claude project set up for my resell business, so it knows what we're trying to do. I start a chat inside that project and say:
Hey Claude, I would like to build out a knowledge base for an AI customer email support agent. Can you create a knowledge base on our wholesale Lululemon program? We have more and more people interested in this program and I cannot keep up with all the emails coming in for people who want to know more about the bundles that we offer. Can you create the knowledge base so I could give to my AI agent?
I paste what Claude writes into a simple Google Doc, then brain dump my own notes on top: what we offer, the questions I keep answering, and details like which brands we carry (primarily Lululemon, with Skims and Vuori in the future). Some of it will need editing later, but I keep it simple.
Back in Lindy, I connect my Gmail account, then add the doc through the Google Drive option in the knowledge base step. Give it a minute or two to fully sync before you click save. Knowledge bases can be big, too. I've built one from a year's worth of Zoom calls from our reselling coaching program, and it could still answer a lot of common questions.
What broke on the first test and how I fixed it
I deployed the agent and sent a test email asking about bulk Lululemon, pricing, starting with 10 items, custom options, and payment. You can watch the agent work in the tasks tab. Mine failed with this error:
failed to find output reference email subject because no source block found
I got ahead of myself and thought I could one shot it. Here's what I changed:
- Trigger filters. On the email received step, I added conditions so the body has to contain Lululemon, bulk, or wholesale.
- Reply step. The reply was set to use Claude 4 Sonnet. I checked the recipient field, added my name as the signature, and saw there's an option to ask me for confirmation before it sends, plus a way to CC people.
- Context. In the settings tab I added: You are a customer support agent and your job is to answer customer inquiries about our Lululemon wholesale program. If you do not know, you are to escalate via Slack DM.
- Sync. The knowledge base hadn't finished syncing, which was part of why I was getting errors.
- Paste the error back. I copied the error into Lindy's agent builder. It said the output references were incorrect and fixed the email fields. I also updated the from and subject fields in the reply step myself, then tested and deployed again.
The replies the agent sent
After the fixes, I got a reply within a minute. It thanked me for my interest, confirmed that 10 items is fine to start, listed the 10 pack starter bundle at $350 plus shipping, and explained that custom selections are available for orders over 25 units. It didn't clearly tell me the next step, so I sent another test from a different email asking what's included, who pays for shipping, and what to do next. That reply said the buyer pays shipping and told me to reply confirming the 10 pack starter, after which I'd get a quote and payment instructions, with items shipped within 24 hours and tracking provided.
A reply inside a minute feels a little fast to me, so I'll probably add a delay of a few minutes so it reads less like a bot. I expect this to cut at least 5 to 10 emails a day, and I can point people who DM me to email instead.
Key takeaways
- Lindy's customer support email agent template plus a plain English prompt builds the flow in under two minutes.
- The knowledge base does the real work. I drafted mine with a Claude project and added my own notes in a Google Doc.
- Let the knowledge base finish syncing before you test, or you'll chase errors that aren't really there.
- When a test fails, paste the exact error back into the agent builder and let it fix the flow.
- Use a delay or the confirm before sending option if you don't want instant, fully automatic replies.
FAQ
Can Lindy AI reply to customer emails automatically?
Yes. In the video I set up Lindy's customer support email agent template with Gmail, and it searched my knowledge base and replied to my test emails within a minute. You can also add a delay or have it ask for confirmation before sending.
How do I create a knowledge base for an AI email agent?
I asked a Claude project that already knew my business to write one for our wholesale program, pasted it into a Google Doc, and added my own notes and common questions. Then I added the doc in Lindy through Google Drive and waited a minute or two for it to sync.
What happens when the Lindy agent doesn't know the answer?
I set it up so that if no answer is found in the knowledge base, the agent sends me a Slack direct message instead of replying to the customer.
Does Lindy AI integrate with GoHighLevel?
When I recorded this in August 2025, I couldn't find a GoHighLevel integration in Lindy. My workaround was to forward those emails to Gmail so the agent could answer them there.
Transcript
Read the full transcript
0:00We're going to be building an AI agent that responds to all customer emails. I have another business and we've been doing a lot more wholesale with clothing lately and I've been getting a ton of emails that I personally cannot keep up with. So, if you're in the same situation and you're tired of responding to constant emails, especially common questions, then this is going to be the video for you because we're going to be walking you through step by step through the process using Lindy AI to build out this customer support email agent. So, let's get right into it and not waste any more time. They already have a pre-built template, which is why I love using Lindy. Over thousands of templates available, and you can see right here, customer support email agent. We're just going to click it and we're just going to go through the motions. So, what I
0:46like about using Lindy is it's just plugandplay. It is so easy to use. Coming from someone that had no experience in the AI world just months ago to now setting up all these workflows that can work aically, which is beautiful. Hands off. Let it rip. Just give it a knowledge base and it will answer all the common questions that you are running in your business or common problems that you're coming up with and you get emails and you're getting annoyed, then this is going to be the agent that we want to build out. So, I want an agent that will monitor and my support inbox. It will help answer customer questions using a knowledge base. If the agent does not know the answer, then it will just escalate to a Slack direct message. I'll help you create a customer support agent that monitors your inbox, answers
1:33questions using a knowledge base with Slack escalation when needed. To get started, you need to click on the link to f to the flow editor where we could build the agent together step by step. So, Lindy has been really nice. We have another agent going on and we're going to be building these agents out every week pertaining to my other business that I'm running and really implementing more automation because I don't really have the time to be answering every single email, every single text message, and keeping up with like the landing page, the websites. Um, so we're going to have Lindy just implemented all throughout the business, and I'm going to be taking you step by step through it. So, I'll help you create a customer support agent. Okay, first we need to pick an email uh service provider. I'm just going to use Gmail. Um
2:19knowledgebased content. Do you already have documentation, facts, or other content that should be included in the knowledge base? And then the Slack escalation. We just got to connect our Slack. So, what I'm going to do is I'm going to go over to one of my projects in Claude and I'm going to just have it write up a knowledge base based on our consignment program. All right. Now I'm in my project with Claude and I can just basically just tell it to create a knowledge base. I'm creating a customer support email agent and I need to have a knowledge base based on our consignment program. I want to be able to answer these questions that are very common. Create a knowledge base that I can give to my AI agent. All right, here's the
3:05projects that I have set up. I'm just going to click on my resell business project because this is what knows what we're trying to accomplish the most. Has all the knowledge that it needs to be able to answer the common questions. So, just going to start a new chat within the project. Hey Claude, I would like to build out a knowledge base for a AI customer email support agent. Can you create a knowledge base on our wholesale Lululemon program? We have more and more people interested in this program and I cannot keep up with all the emails coming in for people who want to know more about the bundles that we offer. Can you create the knowledge base so I could give to my AI agent? And just like
3:50that. So while it's doing that for Lindy, we need to get an email provider. And just like Claude, we can talk to it, which is really helpful um when you're building these out. I'm going to be using Gmail and I have the knowledge base ready to go and I can connect my Slack account as well. All right, Claude is building it out. Let's see. Just making sure it's all good. All right, that looks great. All right, now I'm just in a simple Google document. I'm just going to say this Lulu wholesale program. And then I already have some content here on what we offer. And you just feed it. you just feed it. So, gonna go back to this. I'm gonna say
4:37some stuff we will have to edit, but it's important that we just keep it simple. All right. Brands do we carry primary Lululemon, but Skims and Viori in the future. All right. So, I just basically brain dumped a lot of the information on what we're offering as well as what Claude had put in there. So, just gives it a basic understanding uh questions and answers that I've been responding to and it's just been a little bit time consuming getting these responded. So, now I'm going to have this chatbot automatically respond, which is amazing. So, after talking to Lindy, we just said I'm going to be
5:23using Gmail, filling them in. Perfect. They'll create a customer support agent that monitors your Gmail inbox and answers questions using a knowledge base with Slack escalation if needed, which is amazing. When an email is received, Gmail, it will search the knowledge base for the relevant information. If the answer is found, the commission condition will be met and it'll reply to the email with the answer. If no answer is found, then it will just send right to the Slack DM, which is really nice. So, is this correct? Yes, it is. Great. I'll create a customer support agent that monitors your email Gmail box. um search the knowledge base and go on from there. So what I've been trying to do also I use go high level and for some reason I haven't been able to find the integration for go high level yet but I can just have those emails automatically
6:10forwarded to my Gmail so they get responded. So, for now, in the meantime, um that's what I'm going to be doing. Um because I have a CRM as well. And so far, I couldn't find the integration for GoHigh Level. Maybe you know, let me know in the comments down below. But for now, Gmail works in literally less than 2 minutes. This is already built out for you. And if you were to do it by yourself, it would obviously take longer, but you would understand how the flow works, which I do think is important when you're building AI agents out and you're coming from no experience. But nonetheless, you can see here we got it all set, which is beautiful. It looks amazing. So, let's get it going. Uh, the agent will automatically monitor the
6:55Gmail inbox. Okay, so I think we just need to set it all up. So, let's connect our Gmail. All right, I just added my email and let's see. Account connected. Excellent. Your customer surf support agent is now fully configured and ready to use. The agent will obviously monitor the inbox and respond to questions. So, let's see. It does not have my knowledge base yet. So, let's get that going. I like to just connect the Google Drive. See, add Google files should be here. And once the knowledge base is selected, give it like a minute or two to fully sync. And once it's fully synced, you can click save. And then it will use whatever is in your knowledge base to respond to questions. And I have built
7:41knowledge bases with half of the Bible. Like literally I took a year's worth of Zoom calls for our reselling coaching program that we did and just copy pasted everything into the knowledge base and it was still able to answer a lot of common questions when it came to the reseller niche. So it's really cool where technology has come and if you run a business you don't want to be responding to all the common questions that come in. Maybe you have a VA or maybe you're doing, you know, responding at certain times during the day. It's just so draining, especially if it's a common question. Like for me in one of our businesses, like I get the common question, well, how much is it per unit? What is in the bundle? Is it leggings?
8:26Is it tops? Is it bras? So, all the common things that people just don't read, honestly, and all the information is in the email. But then when they ask a follow-up question like that, I can be like, you know, speed to lead, you know, ask that uh they can ask that question and get it answered relatively quickly. So, um it's up to you how fast you want things to be responded. So, if a a customer question comes into my inbox, I can have a like a delay, a couple minutes delay, 10 minutes delay, so it's not so automated and seems like a you know, a bot that's on the other side. After it is synced, we can run a test just to see if this works. Um, so we are going to test it. Oh, let's just deploy
9:12it and try it out. I haven't connected my Slack account yet, so I should probably do that. Oh, it's already connected. Okay, so it is deployed. So, let's go ahead and start a new All right, let's start a new email to CPSbulk wholesalegmail.com [Music] Lululemon bulk. Hey, I saw your email about offering bulk Lululemon. What are the details? If I just wanted to start with 10 items, is that possible? How much does it cost? Also, what's included in these bundles? Do you offer custom options and how do I make payment? All right, so it's pretty common question.
9:58We are going to send that in and watch it work. So you click on the tasks button and then you can see it do its thing here. All right, so it already came in. It's already reading it. Hey, I saw your email. Blah blah blah. Searching knowledge base. Probably gonna adjust the time in which it responds. I feel like getting a response within the minute is a little uh I don't know just delay it a little bit for me in my end. I think that would work a little bit better. Okay, so we kind of got ahead of ourselves and we didn't set up the entire flow right. So it says failed to find output reference email subject because no source block found. So let's go back and get this all figured out. So email received search the knowledge base. So email received uh filter events. Okay. So, property to
10:46filter on contains. Let's just put Lululemon. Lululemon condition. Okay, that looks right to maybe subject. Let's just put in the body. And let's add another condition. Uh, let's see. Subject, let's do the body contains bulk. And then let's do another one. Body contains wholesale. That's probably filter by email address type. Okay, that looks good. Let's see if it saved that. Okay, that's cool. All right, so we didn't we didn't set this up all the way. We just thought we could oneshot it. All right, so answer found looks
11:31good to me. Let's just delete that condition and let's um deploy this. Oh, hold on. Answer found reply Claude for sonnet is the model that we're using. So you can ask for confirmation before it sends the email to confirm with you is if that's what you want it to send which can be helpful to required the email address of the customer who sent the original email. Looks good. You can also add uh you could also CC people and all right cool. Um I don't the signature will be sent via Lindy. Um, I'll just put my name, Mo Johnson. Cool. Now, let's deploy that and try it
12:16again. Hey, I'm interested in the Lululemon you have to offer. I want to know price points. Can I just buy 10 to start? What's included in some of the bundles? And am I able to order a custom bundle if I wanted to? What are some other brands that you have? I'm interested in mainly Lululemon. I just want to know what items are included in the bundle packs. Send that out. Give it a minute. Also, another thing I found helpful is when you go into the settings tab, you're able to kind of give some context to the agent you're trying to build. You can also include a greeting message, which I'm not going to include, but uh so context. You are a customer support agent and your job is to answer
13:01customer inquiries about our Lululemon wholesale program. If you do not know, you are to escalate the via Slack DM. And let's see, that should automatically save. Yep. All right. And tasks save. All right. So I just reforwarded the original email and now let's see if it can work through it with all the connections configured. So the thing that came up was it didn't finish syncing our original knowledge base which is why we were getting these errors here. So once I officially synced it up, it now knows what to do. Another thing too when you're testing out the workflow to make sure it works according to your preferences if you come up with
13:48errors you could just literally copy paste it back into the agent builder and I did that an error occurred failed to find output reference email.s subject because no source block found and it says I see the issue. The output references are incorrect. Let me fix the email fields in the references workflow. I can see that the agent has been configured and is working well. The workflow is set up correctly and the triggers work. The agent is now fully functional and ready to handle your Lululemon wholesale inquiries. So, let's rerun the test again. Let's go back. Email received. Let's go with this one. Maybe I got to deploy it now. Let's run a test. All right. So, I'm just testing the workflow again after kind of figuring out why it was not working. And
14:35basically, um, the reply with answer was not set up correctly by the agent. I put email from email from for two and the subject. seemed like it had a hard time figuring that one out, but I just updated it, tested it, deployed it, and let's check if I got my email response. All right, so I just got the email within a minute. It says, uh, subject is just responding to the subject line. Hi there. Thanks for your interest in our Lululemon wholesale program. Excited to help you get started. Yes, 10 items is perfect to start with. 10 pack starter bundle, $350 plus shipping. Perfect for new resellers or testing new markets. Shows you what you get. Making sure it's 10. All right.
15:22Why choose our program? Custom options. Bundles are pre-curated for optimal value, but custom selections are available for orders over 25 units. For your 10 pack starter, you'll get our expertly curated mix that's proven to sell well. I love it. Size distribution. Oh, this is great. This is great. Then I can reply. Right on. I'm ready to make a purchase for the 10 item starter pack. Can you help me take the next step? What do I need to do? Even though I think it tells me uh let's see questions. Okay. I mean, that's didn't really tell me like, hey, do this for the next step, but let's see what happens. Let's go
16:07back to tasks. Maybe that wasn't it. It's this one. All right. Just to verify, I sent another email just to really make sure this thing is working. I sent it from another email and let's see what it comes up with. I basically just asked, "Hey, I'm interested in the 10piece starter pack for Lululemon. I want to buy some in bulk. What's included in the 10 pack? Who pays for shipping? What is the next step? I have money ready to go. What should I do?" And I already got an email back. Response: Thank you for the interest in the Lumin wholesale program. I'm excited to get you started. 10 pack shows included. You pay for shipping, fast fulfillment. Next step to get started. Since you mentioned you have money ready, here's exactly what to do. Reply to this email confirming you want the 10 peak starter pack. I'll send you a quote and payment instructions.
16:52Payment confirmation. Items curated and shipped within 24 hours. Tracking provided. Awesome. Really cool. Okay. Now, let me respond back. All right. All right. So, after playing around with it for a bit, you can see I've made multiple inquiries and it's returned successful results. Really happy with the responses and at the very least this is going to cut down at least five 10 emails a day or constant direct messages a day and I can just refer them to email me and it's amazing. So, if you have a small business and you're tired of juggling customer inquiries, you're going to want to check out lindy.ai. I put a link down below and check it
Ready to start building with AI?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolKeep reading
Lindy AI Chatbot Tutorial: Add a Support Agent to Your Website
I use Lindy AI to build a customer support chatbot for easyflip.ai from one prompt, give it a Google Doc knowledge base written by Claude, set up Slack escalation, and have Claude Code add the embed script to my landing page.
This Restaurant Cut Response Times From 4 Hours to 2 Minutes With One AI Agent
A restaurant was taking four hours to respond to customer messages. They deployed one OpenClaw agent. Response time dropped to under two minutes. 80% of inquiries now handled automatically.
Nat Eliason's AI Agent Felix Made $14,718 in 3 Weeks From $1,000
Nat Eliason gave his OpenClaw agent Felix $1,000 and said go build a business. Three weeks later, Felix had generated over $14,718 in revenue. No employees. No freelancers. Just one AI agent on a Mac Mini.
