Deep dive into AI-powered scripting: Mastering IT automation

Webinar transcript

Ariel: Okay, let’s start. Welcome to today’s webinar which is about AI-powered scripting. Now, we have webinars about basically everything within Atera and outside of Atera, including partners, add-ons, security, backup—whatever it is. Right now, we have this amazing AI webinar series, and today we’re going to be talking about scripting in particular. So, my name is Ariel and I am a co-host together with Emanuel. Say hi, Emanuel.

Emanuel: Nice to meet you, everyone. 

Ariel: Okay, and on questions and answers (Q&A), we have Lior. 

Lior: Hi, everyone. 

Ariel: So, this means that we’ll have a full schedule today. We’ll start with a short introduction about Atera and about AI within Atera, and AI in general in the industry. We’re going to go over a demo which Emanuel is going to lead, covering a lot of different use cases using AI to help your scripting and really customize the actions that you can do through AI. Then, at the end, we’ll also have a questions and answers section. Throughout this entire webinar, you can actually ask questions in the Q&A tab. If you see, you have a chat tab and then a Q&A tab. Any questions that you might have throughout the webinar, please write them in the Q&A tab and Lior will answer them throughout the webinar as well. Just so you know, a recording of this webinar will be sent to you after it is finished, so you can have it and watch it whenever you need. Like I mentioned earlier, you can write any questions you would like about AI and AI scripting in the Q&A tab and Lior will answer your questions as we continue. By the end of this webinar, we’ll also launch a really, really short survey just to get your feedback and to make sure that these webinars continue to keep the highest standard.

My first tip of the day is to take out your smartphones, scan this QR code, and join our community. It’s a really good community where you can ask any questions you’d like, both about Atera and the IT industry in general. You can get help about CVEs or the next update from Microsoft or whatever it is. Ask questions there and connect with other IT professionals, so I really recommend joining the community. 

Now, just a little bit in general about AI within Atera. AI in Atera really touches basically every part of Atera—from ticket generation to knowledge-based articles created within a few seconds to diagnostics on devices and much, much more. You can solve tickets automatically with AI, so you can perform a large variety of actions using AI. But today, we’re only going to touch on one small part of this, which is script generation. Even though this is just one part of the entire feature in the toolkit, it is very, very powerful to be able to create and launch your scripts within seconds using AI. 

First of all, why use AI for scripting? As you must know, you can spend loads of hours creating your own scripts. Basically, everyone that I’ve spoken to in the IT industry in the last few months has been using AI to help with scripting. You can reduce errors and even fix errors—we’ll go over this in a bit. You can also provide a customized response to specific issues a lot easier. At Atera, we really believe that any IT management company or IT manager not using AI will fall behind. We believe that this is a tool that’s here to stay, and it’s better to use it and progress with the technology than to refuse to use it and fall behind.

Just a few of the most common scripts that people have created within Atera—we’re going to touch on some of these throughout the demo. As you can see, they range from pretty simple scripts to more complicated scripts. You can really just customize and play around with them however you’d like. 

A lot of people ask, “Why should I use the Atera AI co-pilot for scripting when I can just do it with ChatGPT or whatever AI engine I prefer?” Well, first and foremost, it is because the AI co-pilot within Atera is really trained on IT knowledge, and this is what it does best. It’s targeted knowledge. Of course, we also have it integrated within the Atera workflow, and you can also test it within Atera. We’ll see that in a bit. You can save your scripts and deploy them immediately. It’s just a timesaver—that’s the biggest thing. You’re saving time creating scripts and launching them from within Atera, and therefore, also saving your business money. 

Now, we’re going to go to our live demo. Emanuel will take over and show you a few different parts of the system. We prepared a few different examples for you. Again, feel free to ask questions throughout this demo. Emanuel, please take over. 

Emanuel: Certainly. So, I’m going to do a brief demo of our scripting module and how it integrates with AI. I will also additionally show some automation capabilities, but I will not go too in-depth into that. Let’s start with the bird’s eye view here. This is your script repository—this is the private script repository in your account. In addition to this, you also have a shared script library. These are basically scripts that have been shared by other Atera users. This is not a forum; this is available directly within the application. We obviously validate the scripts for any potentially malicious content. If we find anything suspicious, we block them. So, there is an approval flow to these scripts being posted here. That being said, keep in mind that some of these are environment-specific, so there might be some dependencies. They might not work out of the box; some will require tweaking, but most of them should do the trick for you. But let’s just say you need a script that you cannot find within the shared library. This takes us to our script editor. Obviously, like an editor, you can write your script here. You can even specify certain parameters such as the max run time. This is basically a timeout value you define for how long the script will run before a timeout will occur. Let’s say it’s a very simple script—you can keep it under 10 minutes. Maybe it’s a bit of a complex script—you need to download some files from a shared drive, run some installers, perform some validations—in which case you can expect it to run a bit longer. You can simply change the timeout value to up to 60 minutes. We also support arguments and the ability to run it as the system account or as the current user. But let’s dive into the actual AI stuff. So, what can you do with AI? This is the very first section where we integrated AI over two years ago. It started out as a very basic generative AI—the same thing you would do in ChatGPT—but over the years, we have improved it and added some additional functions which I will cover.

Emanuel: So, let’s start with a common use case. I need a script to clean up the temp folder in C:\Windows. Pretty basic, right? 

Ariel: And we spoke earlier, Emanuel, and you mentioned something that I really think is good to mention to everyone: you need to be quite specific when it comes to script prompting, correct? 

Emanuel: Yes, yes. Obviously, there is a degree of prompt engineering required whenever you’re working with AI. They are able to understand context most of the time, but the more specific you are, the better it is. So, in this example, I just gave it the exact path C:\Windows. Although it would have probably understood “clean up the temp folder in C,” it might have asked you to specify the path at some point. 

Ariel: Okay. Now, this takes me to the second advantage of using the scripting in AI: you can actually run live troubleshooting by assigning a live agent and testing the script itself, and you’ll be able to see the output. 

Emanuel: There we go. But before running this, I actually want to run a script to check if the temp folder needs a cleanup first. I’m going to save the script. Let’s give the category “maintenance” and create that. So now, I want to see what is the size of the temp folder in C:\Windows in gigabytes. Let’s see here.

Ariel: What’s cool with this function is really the amount of time that it saves technicians. Otherwise, I’ve seen it myself—people write their scripts on ChatGPT, then they have to go into Atera, save their scripts, copy them, save them, launch them, and try them. And then if there’s an error, they have to go back again to ChatGPT, recreate it, resave it, etc. But here, they can do everything just in one window. 

Emanuel: Yeah, so this time it’s taking a bit long, and I didn’t actually catch this initially, but it’s using a faulty method. It’s basically running a loop measuring each object, which I don’t like because it’s ineffective. So, let’s say “read the size from properties.” You can actually modify the script instead of rewriting it from scratch. Now, let’s change the device. There we go, a lot faster, much more efficient—6.5 gigabytes. Quite a lot. [Laughter] So, this is a problem. Let’s save this here and try to run the cleanup script that we just created. 

Ariel: Okay, some errors here—some file access paths which are not accessible. Now we’re running the script that we don’t know what happened. So, this takes me to another feature which we offer, which is the inline editor through the AI. I want to see the temp folder size after the cleanup is finished. 

Emanuel: The inline editor here—basically, if you type in a pound sign or a hashtag sign, then you can just edit and continue the AI generation of the script within the script and just add on continuously. Okay, so it’s down to 2.7 gigabytes. There’s still some files which it couldn’t access—probably some leftovers from installers which the system does not have access to. I don’t want to see this output; it’s a bit messy. So, I also want to do something else like suppress any errors related to file access. I think I need to run this a bit in the beginning, but I don’t want to waste too much time troubleshooting this. 

Ariel: Yeah, I think let’s continue. I think this is actually a great example that AI is a fantastic tool, but you can still make mistakes with AI. This is my error, for example. You still have to put in the time and actually check that it works. One huge recommendation is to check the scripts that you create with AI—or any script for that matter—in a sandbox environment or on a test device. Definitely do not run it on all your devices without testing it. 

Emanuel: Especially if it’s something that can affect production environments. I want to briefly go over some of the automations that we can achieve with these scripts. I’m going to move to our alerting profiles, which we call thresholds. I’m going to add a profile called “temp cleanup.” We’ve got a few industry-standard presets here: personal computer, Mac, laptop, Linux as well. But what I want to highlight here are the script-based thresholds. I have the list of defaults here which you can select from, but let’s say you want to monitor an item which is not specified here—in our case, the size of the temp folder in C:\Windows. 

We’re going to do it through a script. I’m going to make it a simple warning. We’re going to use a numeral threshold, and let’s say if the value exceeds one for a duration of more than 15 minutes, we want it to automatically run the cleanup. 

Ariel: Okay, so before you exit this, I just want to go over this really simply. Thresholds are items or threshold items that you can create in order to get alerts in your system about basically anything that happens within your network. We have a long list of presets that you can just click on and create your own alerts very quickly. But if you want something very specific, if you want to monitor something very specific, then you can just create your own monitoring script and you can also automatically launch an auto-healing script which is what Emanuel just added. Meaning anytime that threshold is met—in this case, anytime the temporary folder has more than 2 gigabytes over a one and a half minute period—then automatically launch the script in order to try to remediate the situation. 

Emanuel: In addition to this, if the size is brought below the threshold value of 2 gigabytes, the alert will self-resolve. This feature is currently supported on Windows and Mac devices. We will likely very soon expand it to Linux as well. 

Coming back to the scripts, I want to highlight another feature that we offer, namely the ATA variables. Let’s say you need a script that needs to perform an action either using some credentials or by specifying a value that always changes—it will never be the same. So, let’s say something simple like changing the computer name of the device and initiating a reboot. Now, the AI is also feature-aware, meaning that instead of giving you a standard variable and just some general text like “please specify the value,” it will actually use the ATA-defined variables. How does this work? Basically, once you create the script, you will go to a device. I should have checked the name—change device name. Computer—change computer name and reboot. Yeah, this is the one. So, instead of simply running this script, when you hit run, it will actually prompt you to enter the value. Very useful if you need to pass some credentials which you cannot simply store in plaintext in the script, or any other use cases where the variables change. One of the most common use cases for this is when someone wants to create an Active Directory user and they have a script for that. They just have these variables where they specify the name, the manager, and also the name of the security groups where it should be added, and the script handles this automatically. 

Another automation feature that we offer in ATA is for the automation profiles, where you can basically schedule tasks to run at a certain time interval. Let’s say you want to create a maintenance script that instead of having it run based on a trigger like in the example with the temp folder exceeding 2 GB, you just want it to run periodically. So, in this case, I’m going to go to the patch management and automation profiles. I’m going to create the profile. I’m simply going to attach the script which cleans up the temp folder. Obviously, in this case, we also have it here, but there’s many more things you can automate through scripts than just out-of-the-box options. I’m going to add a schedule, set the frequency. Let’s say it’s going to be monthly on the first day, on the first weekday every month. Let’s say 9:00 PM—people should be home by then—and run it. Also, we do have a retention feature here which just in case the device happens to be offline and misses the schedule when it’s executed, we do have a way to set up a retention period.

So, let’s say in this case, if within a week of the schedule, the device comes back online, it will immediately receive this automation task and do the cleanup. 

Ariel: Excuse me, Emanuel. Before we move on, I wanted to try to understand who is watching right now just to see the level of scripting experience. So, I’m going to launch a quick poll before you move on to the next features. This will help us understand if people are complete beginners or if they have some scripting experience, etc. 

I’ve launched the poll, and it opens up another tab next to the chat and Q&A section. If you could please just vote there—it’s just a one-click vote—so we can understand who we’re talking to, how much experience you have, and whether this is too high level or too low level. 

Right now, it seems like most people have a little bit of experience, but very few advanced scriptors out there. So, it seems like we’re on the right track, Emanuel. We can keep it pretty basic, but there are a few others that are a bit more advanced as well. 

Emanuel: Okay, so this is actually an excellent use case for the scripting co-pilot because it can help you within safe parameters to learn scripting, really bridging that knowledge gap. 

Ariel: Sure, may I proceed? 

Emanuel: Yeah, let’s. I’ve closed the poll; people voted, and now let’s move on. 

Perfect. There are two more features I would like to highlight here. Again, they are tangential to scripting but do use scripts. Let’s go and have a look into a device. From a bird’s eye view, you can get the device name, domain, IP, some hardware information, disk information, OS, and security information. But let’s say you need to monitor some specific parameters which are not available out of the box in the device interface. How do you do this? Well, for this, we have implemented script-based custom fields on the device. I’m going to show them right now. We support custom fields on basically every single object in Atera, but let’s focus on the agent for a bit because this one’s special. We have the normal text, number, date, checkbox, dropdown, dropdown dependencies types of fields, and we also have the script-based custom field. What does this mean? Basically, you can attach a scripted query to this, and the output of that will dynamically populate the field on the agent. Now that value can be seen on the device page, it can be filtered through advanced filters, and also through advanced reports—the advanced reports being a BI tool we have integrated into Atera. In this case, let’s have a look at verifying the TPM status—a very simple script—and also the device uptime. How does this work? If you go into a device and have a look at the custom fields here, you will see that this is the output from a script. This is also the output from a script—same with the uptime—as well as all of the information found here. You also have your classic custom fields, which is basically static data. 

Last but not least, this is something that’s not yet out in production; it’s currently with our design partners. Actually, the co-pilot is out, but it’s a paid add-on, and the auto-pilot will be released, I think, at the end of April if I’m not mistaken. The co-pilot is essentially the assistant for the technician, and the auto-pilot is the AI assistant for the end user. 

Now, what does this mean? Here, with this toggle, you can enable certain scripts to be used by the AI just in case it’s not able to perform an action or a verification that’s not already trained with. So, let’s say I want to see the BitLocker status on the device. How can I do that? 

Ariel: So now, Emanuel is basically chatting with the AI co-pilot. This is another part of Atera, another part of the co-pilot. You can chat and find out information about the agent on the device. What’s really interesting about this agentic AI is that it’s updated all the time. It’s not some random information that you find online on ChatGPT or Perplexity or whatever AI model, but rather this is information coming from your devices. Therefore, it takes a little bit of time sometimes to get these answers—it can take a few seconds; it’s not instantaneous—but it’s your information about your devices wherever they are, as long as they have your Atera agent on them. 

Emanuel: Just to give some context here, we support around 900 automated actions, validations, and verifications at this point. But I specifically asked for something that I know it cannot do. Why did I do this? Because I can make it do it through a script. So, let’s say I wanted to be able to check the BitLocker status, and I also wanted to be able to create a test event. Now, it’s a very simple, harmless script—just creates a custom event in the Event Viewer. I use it for testing purposes when it comes to the threshold items, if you remember in the beginning where we created the custom alert. You can also do it based on Event Viewer IDs. Let’s go back. Emanuel: As you can see, it’s running diagnostics right now, and we got the answer. Amazing. 

Ariel: This is huge. We see a very large percentage of Atera customers use the AI co-pilot. It really just saves time. We have customers that have been able to use the AI co-pilot to assist them, and therefore they don’t have to hire more people. They can cut their ticket response time, ticket solution time, and it’s a really, really big help. Of course, with these scripts, you can actually give the AI co-pilot even more power by enabling it to run these scripts that you’ve told it that it can run. 

Emanuel: And as you can see, just like we mentioned earlier, it can take some time, and this is normal. We want this information—it’s not just information pulled from the internet, but it’s actually actions that you can take on the device using the AI. I think this is not going to work… Oh, there we go! It took a while, but it created the event. This will be much faster once it’s ready for production. And without further ado, that’s all folks. 

Ariel: Okay, so I think we can sort of summarize what you showed now with the scripts as a way to customize features throughout Atera. Whether it’s an IT automation profile that automates tasks and you don’t have that task from within Atera, you can customize that task by creating a script. Same thing with a threshold—if you want alerts for something that doesn’t come from Atera out of the box, you can customize that threshold profile and get specific alerts for that specific threshold that you’ve created. And same thing with the AI co-pilot. 

Just before we summarize a bit more, I am interested to see if we have any co-pilot users here. So, I just launched a little poll to see the percentages of current users and maybe potential users. Just to see how many of you are using the AI co-pilot, how many are not yet, maybe you’re thinking about it, maybe you say AI is not for you. So please vote. 

We have a few votes in. Okay, fantastic. It’s a little bit mixed. I’m going to close that poll. And then just before we leave today, I’d like to see, Lior, do we have any questions that the audience has asked? 

Lior: Hi, everyone. So, I have one I can share which says, “Have the scripts in our shared script library been tested?” The shared script library is a place where we make sure that all the scripts are safe to use and don’t contain malware, but they’re not tested for functionality. So, you will have to test it out on your own to make sure that they work as intended. That’s a question I see is pretty popular and applies to everyone. 

Emanuel: Just to add to this, the large majority will work because they’re very simple scripts for common use cases. Some of them do have dependencies based on system variables or specific implementations of the infrastructure, so always test them in a test environment first. 

Ariel: Amazing. I think Karen asked here in the chat section, “Where is the shared library?” Emanuel, could you maybe just show that again? 

Emanuel: Just a sec. Can everyone see my screen? Okay, so in the admin tab, it’s under monitoring automation, but you can always use the search here—it’s a lifesaver. If you go to scripts, you will see two tabs. This one is your own repository, and this one is the shared library. You can sort them by votes, by author. The best way to go about it is to sort by clones and by votes. These are the most downloaded scripts, the most used, and the most highly upvoted. 

Ariel: Great. Any other questions? 

Lior: I think that’s it. If we failed to answer any of your questions throughout this webinar, we will answer them once this webinar is done. Karen, the recording will be sent to all of you once this is finished. 

Ariel: I am going to launch a short survey because it’s just really important for us to get your feedback. So please, please answer—it’s really good for us to improve the webinars constantly and see what you thought about this webinar. 

We really appreciate you joining today, and we hope you found this educational and interesting and that you will join future webinars as well. Like I mentioned in the beginning, we have webinars every single week, and now we have quite a lot of AI webinars—almost one every single week on a different feature within the AI co-pilot. You can join onboarding webinars where you can learn Atera from scratch, and we also have a lot of on-demand webinars on our website. So feel free to join both the on-demand and the live webinars. 

Thank you for joining today. We really appreciate it. We know you have very little time for non-ticketing urgent matters, so we appreciate you taking the time. Thank you. 

Everyone: Thank you, everyone. Bye-bye. Cheers. Bye-bye.