Advanced Workflow Automation with AI

[Intro]

What if your support system could handle not just simple questions but even your second level tickets all without lifting a finger? Sounds too good to be true, right? Well, today we're diving into advanced support automation with AI. Exciting. We are talking about automating refounds, trial extensions, bug reports, and even gathering extra content to make every conversation smarter. Let's jump in. All right. So, you've mastered the basic of AI in support and you've got your AI assistant answering common questions, pulling data from your knowledge base and recognizing customer intents. But now it's time to level up because while automating FIQ is great, those second level tickets, refounds, trial extension, back reports are what really eat up your team's time.

The good news with intense missions and rest API we can automate those two and by the end of this lesson you will have a solid plan for taking these workflows from manual to magical and this is the reaction Carola had talking about how they handle those secondary tickets in seconds instead of weeks. She's power.

[Why Automating Second-Level Tickets Matters]

So, as you can see, this is so important. And second level tickets are the ones that usually requires a little more effort. Collecting customer data for a refund request, you know, back and forth to gathering data, verifying details before extending a trial, submitting a detail report back. But here's the problem. These tasks can pile up, slow down your team, and frustrate your customers. And sure, you can assign this conversation to a human, but why not let your AI handle it instead? With advanced automation, Aura, your assistant, can do the heavy lifting like asking the right questions, gathering data, and even triggering actions through a rest API.

Therefore, your team spend less time on repetitive tasks and more time on the stuff that actually matters. So, here's the game plan for today. Okay, we will look at four advanced workflows you can automate with AI. First, refund request, trial extensions, but reapers and asking follow-up question for additional context. For each one, we will go over how it works, how to set it up, and how to make sure it's running smoothly. Plus, I'll show you some chat flows templates to get you started. Let's start with refunds. Okay, refunds are perfect example of a task that's repetitive but require precision. Here's how you can automate it.

Step one, let's

[AI Intent recognition intro]

use the intent recognition type of job flow. When a customer reach out saying I would like a refund, the AI assistant detects the intent refound request. Step

[AI Missions intro]

two, this will launch a mission to collect data. The mission is where the assistant ask the customer for all the information your team needs like plan type, invoice number, last four digits of payments and that would process basically the data of your customer. Step three, trigger an action with rest

[Use AI + REST API to act on support requests]

API. So once the data is collected, the assistant uses a rest API to send it to your payment processor, initiating the refund automatically for you. No back and forth emails, no manual steps. Everything is just working in a seamless process from start to finish. Let's see how it works. We created a template for

[Automate Refund requests with AI]

you. So you can just click over here and click on the refound. Now if you click on the intent, you're going to see that the description is quite easy. The user wants a refound. I want a refund. I've been build twice this month. You can you can provide different matching intent examples if you see that your customers are asking for a refund in a different way. Now we're going to collect some data over here with AI mission. Let's say you need to understand if the user is eligible for the refund and the user to provide the email you sent the invoice to the amount of the refund and amount of invoice.

So by prompting the nishian this way and by adding this conversation properties now we can actually send this data to your back end check if they're eligible for it and then with a response we're going to tell the customer if you know the the refund could be process it or not. So this is the mission. It's fairly simple. Now, instead of assigning this, of course, you can assign it to any of your team whenever the mission finished to collect the data for you. But we're going to do way much more. So, we're going to use a rest API over here. And let's say you already got your dedicated rest API to understand if the customer is eligible, something like API company.co.

Now, you can add of course some environment variables. Now you can also authenticate those rest API of course and you can save the API keys in one of your end variables under your project settings. So once you got that you can just record it this way. Now let's say I want to pass all the information I've been collecting with Aura and now I can customize the full body. For example, invoice email and then I want to recall it over here. Um invoice invoice email. Cool. Now um I can pass anything over here that I already connected. Now I can also capture the response. Okay. And uh this is the way to pass the data from your back end to uh to aura as well and giving it under a conversation property.

So this is of course is this is extracting the data with a JSON notation and store it into a property. So let's say your data is going to be uh something like uh this just for the sake of this. So let's say your payload of the response is something like this. You will need to do response doeditable and you can store it into a property. You can add a new property a conversation property refound editable. Now you can store it under trles can create the property. That will be it. Now you can do the same with reason response dot reason. And this would be again another conversation property refound response reason.

This would be a text and that would be it. Now we don't need this anymore. Cool. So we're sending out to you to your back end uh the invoice email you collected with Aura and now we're expecting back a response eligible uh through vaults and a response reason with a response reason over here. Now when there is a success of course you can pass this information again to another mission giving it the con the context of what you um did receive for example the reason and uh as well the refund eligible and now here in the mission goal we're going to say check if the customer is eligible to eligible for refound And if not tell them the reason otherwise tell them they will receive the money in five 7 business.

That's it. This is how you would do it. And uh basically if there is a failure you can add like a quick message uh something went wrong or you can just escalate it to to your team. So you can get rid of this. So if your back end is failing, you can of course assign this to uh your support team. And here we go. So if instead the second mission goes okay goes into success now you can just close the conversation directly or if there is a problem with this now you can escalate again to one of your team or just launch the chat flow we created before to escalate directly based on the customer type.

Now again over here we need to define the escalation to the very specific team or again we're going to use the very same logic for the escalation. So we're going to launch the chat flow again for the escalation round. This would be it. This is the way you can fully automate a refound with uh the help of the intent the AI mission send it to your rest API collecting the data providing the response to the customer and close the conversation directly. So now we're going to publish this. Now of course this REST API will not work. So it's going to be a failure. But we're going to see how this is working.

Okay. Now let's preview this and see how it's working. Now I want to start a conversation asking I want to refund. Let's see how Aura behaves on this. Now she's going to start typing then stop for a sec. This is the the signal that she got into the intent of the refund. It's cool. So, she got it. Now, she can help with that. Could you please provide the email used to send the invoice to the amount, the amount was charged, and the amount of the invoice? So, I used my email l.com. It was 99 and it was uh November Cool. I provided all the information.

Now, let's see what is going on in the inboxes under her. So, as you can see, she started the process. She sent the request directly. Now, of course, this is a fake uh request. So, she behaved the exact way we wanted. So, the chat would escalate directly to the customer support instead of doing and going ahead with the other mission. Otherwise, if you had this kind of API endpoint on your site, this would behave exactly the way you um you build it. This is the way you can build a refund. Now, let's talk about trial

[Automate Trial extensions requests with AI]

extensions. And here's the challenge. Customers often request extensions at the last minute and process them manually and slow down everything. But with AI automation, it's easy. Step one,

[Automate Bug Reports requests with AI]

recognize the intent. Let's go back to chat flows. Now we already have a functioning template over here. If you go and click trial extension, you will find something like this. So again, we'll click over here and check the the intent. The customer wants to extend the trial period and the matching intent might be something like this. Okay, that's great. So let's check the mission. The mission uh prompt would be something like to extend the trial. We need to understand the reason why. uh what they're going to try during the extended periods, which plan are they interested to and how many days they need. If they want, you can of course u get rid of this wire.

So, you can customize it. But this is something that um could work. Uh if they want to try the pro features, the maximum days are 30 days. If they want to try the enterprise, it can extend maximum 45 days. Now, um of course, we're going to collect some conversation properties over here. Some of them might be required, some of them are not required. So in this case um all of those will be conversation properties and this conversation properties will be also used in the rest API. Now the rest API is going to uh check against um your your customer ID or subscription ID and check if they're eligible for another extension.

Let's say uh you allow an extension every 3 or 6 months or something like that. you can provide them um and based on the data for example we're passing on here the company ID this is something that we store automatically for you um so if you already set up the company and the company ID um within customer league you would find this um automatically and the way you do it is just typing down company ID or of course if you use a different uh identifier for um this reason uh like stripe customer ID or something like that it's okay you can customize whatever you're using over here.

And by the way, we do have also templates for REST API. Now, uh again, we're going to capture the response over here. And we assume that uh the response of the message would be data results and message. Let's say the message would just tell them uh if the extension was okay or not okay. And then what we are going to do is um communicate in them that your trial have been extended for another 14 days. Otherwise if of course the the rest API goes into failure such as 400 500 that kind of errors uh it will go into failures and um we will say we couldn't extend your trial because and this was the trial extension result message that we've been collecting in the capture response over here.

So again as you can see this has been completely automated from the the very start of the conversation to the very end. Now we can try this out but again the rest API will not uh work in this case because this is fake. But this is working flawlessly for for you. Let's try out uh this flow and ask a question over here. I want to extend my can something like that. So as you can see I already added a few information she's going to ask me because in the mission we defined that we need to provide the try extended reason try for plans and the number of days.

Okay, I understand you want to extend your trial because you were not able to try your program. Could you please confirm which feature you will be trying during my extended period? Again, this is something cool that um you can add or not. Um I want to try your AI. So this is cool. Now if we go under the conversation and see what's going on. Uh as you can see, she fired of course the the intent. She gathered the informations over here. Try extension reason days and triing for AI services. Now she's asking another question. Thank you for information. Please confirm which plan you're interested. The options are startup pro or enterprise.

Now these are properties that I specifically the um told the AI uh in the mission that are completely different of course um those are the the plant types that you can define specifically for your use case. Now I won't try the pro. That's it. So now she sent a request to this API which of course is is not uh is not working and it returns null because of course it was a fake thing. But that was it. That was it. So this is how you can also do try on extensions completely on autopilot with this simple but still um super skilled chat flow uh for you.

Cool. Back reports are another big time drain for support teams. But with a little automation, your assistant can gather all the necessary details before your team even sees the ticket. Okay? So you can use one of our templates such this one for example bugs to Asana and it will be super super easy. Again there will be an intent the user is reporting a bug. You can provide a couple of matching examples. Now the mission is going to ask for more information the critical information. You might need to reproduce the bug for example right? So here you're going to define also at the description level of the conversation properties uh what Aura is going to put into those conversation properties.

Okay. And also the product you can um accept only specific values as we see before. Aura is going to use them if um as a context to to ask for more information and in description as well. you would be providing Aura the kind of products um that she is going to uh work with this ticket. Okay. So um once we got the step now the prompting over here is super simple to better understand the reporting of the bug. Please ask the users to provide the following information. The steps to reproduce the bug the product affected the page where the bug occur. Now again we can use uh this post API to send it to Oana or Click app whatsoever.

You can use your Asana token. Um you can customize the payload and let's see how it goes. So um we already have this. So when uh this is successful, we're going to um actually we're going to just close them. So when this is successful, we're going to assign it to the support team. And when when there is a failure, it's going to be the same. And the escalation would be again the very same output. Okay. Now let's publish this. See how it goes. Now let's see how it goes under aura. Now you see she met the intent. Now she asked me questions regarding I'm sorrying having experiencing issues.

Could you please provide the following information to help us understand the problem better the step to reproduce the bug? The specific product affected conversation blah blah blah. Now, I'm going to provide, of course, all of this information. Something like this. Whenever she completed gathering the information that is needed to report the bug, while she's also asking for confirmation about what I already shared. This is crazy good. Wow. That's cool. So she went ahead sent the rest API with all the information. As you can see, this is the data payload and it was supposed to be like this built with the project is not loading after upgrade notes starts open.

The app is and is stuck. Cool, isn't it? So this is the way to mainly automate the

[Final Thoughts & What’s Next]

majority of the task. Of course, this can be extensive to any other kind of um solutions. These templates will save you hours of setup time and give you a solid starting point to customize for your business. With intense missions and rest APIs, your AI assistant can handle way more than just basic FAQ. As you could see, it can manage refunds, trial extensions, bite reports, and even collecting extra context to pass along to your team. The results, well, your team spends less time on repetitive task. Your customers get faster resolutions and your old support systems run like clockwork. Amazing. That's it for this class on advanced support automation with AI.

If you're ready to take your workflows to the next level, check out the templates I mentioned and start automating those second level tickets for your business. And stay tuned because in the next lesson we will explore how to integrate these workflows into your multi- channelannel support systems. Thanks for watching. If you found this helpful, hit like and subscribe and let's keep building smarter, faster, and support systems together. I'll see you in the next one. Chiao.