Automate customer support tickets with AI (step-by-step Voiceflow tutorial)

Key Takeaways

A team building on voice flow automated 70% of their customer support. It used to take them 30 minutes to respond to a ticket and now using voice flow and AI. It takes zero. I'll show you how you can build this for yourself in less than 5 minutes. Let's get started. Once you've signed up for a voice account, create a new agent. Then give it a name. I'm going to call it support agent, but you could use your business name. And then create it using the basic template. Now, we're going to use AI to generate a bunch of this project because we don't want to have to build it from scratch. So, let's close this. We're going to click on our existing agent and we're going to go and create a new one. So, we'll go here and create agent. Now, I'm going to press this generate button and I'm going to describe what I want to do. So, generate a support agent that collects a person's name, email address, and problem. then create a ticket for that issue. If a person isn't clear, ask clarifying questions first. So now I'll hit generate and the AI will generate us a prompt. We don't need to write that ourself and I'll hit accept. I can make any changes here if I want to, but for now I don't. Now, if I already had a bunch of questions and answers, I could enable access to the knowledge base, and then our agent could try and answer questions first before going and creating a ticket, reducing our team's workload. But for now, I'll leave that turned off. All right. Now, we use Zenesk for customer support. So, I'm going to hit Zenesk, and I'm going to say create a ticket. Awesome. Now, I need to get my Zenesk subdomain. So, I can go get that. Here is our Zenesk instance and I can just go and copy this and paste it in here. Now I have admin on Zenesk and of course I own this project. So I'll now be able to go and connect it up. Zenesk is connected. Awesome. All right. So now let's close out of here and let's try out our agent. So I'm going to test it. Hit out of an issue. My name is Jacine and yep, here is my email address. Okay. What's the problem I'm having? The more issues, the better I can help. I'm going to say my account isn't working. Uh, it wants to give me some more detail. Okay, I can give more detail. Um, oh yeah, I'm getting an error saying it's not secure enough. Awesome. It's going to say just to clarify, you're receiving an error and I'll create a support ticket for you with this information. So, anything else you'd like to add? Uh, oh yeah. uh actually uh the account number is and we could include a requirement for our agent to ask this automatically. All right, so here is our summary. It's going to go and create the ticket and that's it from a user's perspective. They now have an email with that ticket in. But now let's go and check Zenesk. So I'm going to go and open it here. I'm going to go and refresh our Zenesk instance. And here we go. Account security issue. It was me. User is receiving an error saying the account is not secure enough and is prompted to contact us support. Amazing. Now there is one little guardrail I want to add here that currently right now we are not locking the priority. A user is going to say everything is urgent. So we can go and actually modify how our agent uses that Zenesk integration. So let's click here and let's say the priority must always be normal. That way someone's password not working is not going to be urgent and alert our entire team. That would be super annoying. But the rest of these having LLM capture works for me. I could capture a subject too. Um or I could go and have a predefined one. But I think this is fine for now. And that's it. That didn't even take us 5 minutes. Imagine what you could build in an hour. AI is only as powerful as what you let it do, though. So here's how you can let your agent make decisions and build a great customer experience integrating deeply with your business. [Music]