Transcript
Welcome to Autonomy Day. This is probably one of the panels that I'm most excited that we're getting the chance to host. My name is Yoav. I'm the general manager for Eterra here in North America, and today I'm joined by two of my favorite IT leaders, Emerica Lucid and Eric Denholtz, and we're going to be talking about AI and IT and specifically Autopilot and its role kind of in changing the way that IT is done today. So big, big thank you to both of you for joining us today, and kind of before we start rolling with the questions, I would love for you guys just to introduce yourselves, tell us a little bit about your team, and kind of what does your company actually do? Emerica, can you kick us off, please? Definitely. My name is Emerica Lucid. I'm the vice president of technology at CCI. Our company both manages market rate and affordable multifamily buildings across the US, and my team specifically supports the organization in a project management, software implementation, and traditional support capacity. How many people do you have on your team today? My team is 700, and we're supporting about 400 employees. Very nice. Oh, sorry, seven. Seven? Yeah. Seven. Seven people supporting 400. That's a lot of people. Sorry. Seven people supporting 400. Eric, can you introduce yourself, please? Yeah. Hi. I'm Eric Denholtz. I am senior infrastructure architect at Lamps Plus. Lamps Plus is the nation's, in the United States, biggest specialty lighting retailer. I'm on a team of probably about 20 people. There's a few admins, an engineer, me, a director, and then we have about seven help desk technicians, and we're in charge of the whole infrastructure. There's about 1,200 end users, 300-some-odd servers, and a big e-commerce website. Awesome. Eric, let's start with you. Prior to coming across Eterra and starting the conversations, what were the problems that you guys were trying to solve on the IT side? We were trying to solve a lot of problems, you know, all types of problems. Probably the biggest problem of all is being understaffed and having too many tools. I think at the end of the day, that really sums it all up. We were lowering our security stack. We had like four or five different security tools. We had fragmented remote control tools. Our ticketing system was JIRA, which is really more of a project management system, unless you have a specialty consultant design it, which we didn't. We had this tool fragmentation and a lot of inefficiencies. I was actually looking at all different products, but I think when I got to Eterra, I got to solve not just some of the problems that I wanted to solve, but also some other problems that I wasn't even planning on solving. What problems were they? What were the known problems that you came in for? It sounds like fragmentation, lots of tools, kind of a convoluted stack. What else did you know that you need and what were the things that you actually found out that were possible? Well, what's interesting is the actual impetus for me coming to Eterra is really weird. We had a product called Carbon Black App Control, which was a blacklist, whitelist kind of technology often used in financial services and other places. We wanted to get rid of that and we didn't really have an adequate solution to do the same thing. At the same time, we have all these process inefficiencies and we have no good RMM solution, so I'm looking at Eterra on the side for that stuff and I'm having these conversations with you guys and realizing, well, wait, we can actually very quickly and easily basically replace this $50,000 a year product with a simple script via just-in-time access, basically is what we were giving, granting a user access locally as an admin for a very short period of time and then taking it away. So all of a sudden that use case was resolved and then I started really focusing on the RMM stuff, but really the bigger picture of all of it is we had barely any process for almost anything. We had horrible documentation and we were getting our staff cut quite a lot, so we're really understaffed. So I mean, I could talk forever on each one of those points, but Eterra is coming in to help resolve a lot of that stuff right now. Cool. America, can you share a little bit about kind of what was your journey? How did you come across? What were the problems that you were trying to solve? Yeah. In a similar fashion to Eric, we were really just looking for kind of like more of an all-in-one type partner. We also, I mean, it's really burdensome to have to log into seven different systems for our IT support team every morning, then prioritize their tickets and reach out to users and so on and so forth. So we were really looking for that all-in-one platform and we were really looking for also just kind of a more like tech focused partner for the long haul. And so that's kind of where Eterra stood out more specifically with offering the autopilot tool. And so some of the first things that we saw that we were looking for help with was just in collecting information from users. A lot of the issues that we get would come in historically saying, I just, I need help. And so it's like, okay, well, what do you need help with? And so in some of the testing that we did with autopilot, we saw like not only its ability to kind of like diagnose like general IT issues, but really like asking the questions of like, okay, well, like what kind of issue are you having in such a customer friendly way that it's reduced some of the back and forth that our teams have had to do? So yeah, a lot of time saving there. Famously, IT people don't particularly like getting tickets or people reaching out saying things like, my computer is running slowly, right? People are... Not working. Right. My computer is not working, right? Absolutely. So America, when you were kind of evaluating autopilot, kind of what were the biggest factors that you were thinking about? Kind of what were the things that were really part of that decision process for you? Well, of course we wanted to reduce the amount of time that our techs were spending just in kind of the back and forth. We really didn't want to sacrifice the service quality or any of the customer service ratings that we had. We're really friendly with our teams. And so that was a big factor. We were actually really impressed with the out of the box responses that we saw from the agent. And we've actually seen an increase in customer satisfaction from that perspective and customer sentiment. Yeah. They really liked the emojis. The emojis really sell it for them. And then just as critical was the ability to understand our organization's use case. We're in property management, multifamily, our clients are our residents. And so it's really important that any issues that they have are being... If they can't be resolved immediately, that they're getting escalated to the right teams. So my team's fragmented in that way where we have traditional IT support, and then we have a team that's doing project management implementations and even more specific business application support that we use in the industry. And so it was really critical that those kinds of issues, that we did some kind of initial troubleshooting clear browser cache type steps before we escalated it to the team. And that we did that escalation really quickly. And it sounds like right now you've managed to have autopilot do all of that first level stuff back and forth with the clients, taking care of those initial issues. So that when it does get to a technician, they're able to focus on the more high value work. Is that a fair assessment? Yep. Correct. And what was the reaction? Because as you mentioned, you guys have an interesting setup, right? You have kind of the employees of CCI, but there's also the tenants on the other end. Prior to rolling this out, was there any concern about how end users were going to feel having to interact with AI, having to interact with autopilot rather than what they were used to getting up until now? There was a little bit of concern in just making sure that any agent that we exposed to our end user could understand kind of their use case because it is such a nuanced and specialized industry. We obviously don't want anyone to say like, oh, my resident's mad at me. What do I say? And like generate that kind of response. So it already being pre-trained on like IT and IT specific issues was ideal. Obviously you all have some also like risk factors built into that. So those kinds of responses would get escalated as maybe like, I can't help you with this. There's someone that can better answer why we can't help you with this kind of issue. But really, I think the most important thing for us that we learned and that we're continuing to adjust is our knowledge base. So that knowledge base really comes in handy with like industry specific abbreviations that we use with industry specific or system specific knowledge. We do use this application. We might've used this one in the past. We don't use it anymore. So when people come to us, they're really getting the answers that they need. Awesome. Eric, over to you kind of, you mentioned that you had come in for one thing and then through the process and the discussions, which to be completely candid, I was aware of it a part of together with you, but kind of tell me a little bit more about when you heard about Autopilot, what kind of came to mind in terms of the use cases that you wanted it to handle and kind of how is that panning out? Yeah. I mean, one of the, that comes to mind is the first thing America actually said, which is, you know, asking qualifying questions and getting, you know, getting information on, on tickets and speaking back to that inefficiency thing I was talking about earlier. That's probably one of the biggest ones. And simple things like how to contact a person, you know, now is automated into the system. So we're not scrambling to look at where the person is or how to contact them or what their actual problem is. So right away, that's coming up as I'm talking to you guys, but you know, some of the other things that were coming up for me were, you know, upskilling not only our users, but really our technicians. And some of that goes over to the co-pilot side and not as much Autopilot, but a lot of it actually is with an Autopilot because they've gotten more excited about documentation, believe it or not. So, you know, they're seeing that, oh, I can put this information, you know, into this knowledge base and they go and test it themselves. Like if a user asks, how do I find my Cognos username, it gives them the thing they put in the knowledge base. So they get excited about it and now suddenly they're wanting to generate, you know, knowledge based articles and really train it more like America was talking about. And I think that's kind of the phase where now is we haven't fully rolled it out to all the users yet. We're still in the user acceptance phase. We did just get started. So it's not too far in. But we're getting some good feedback for sure. We're getting positive feedback. And I think my suspicion is that when we roll this out, not only will we improve our efficiencies with supporting our end users, I do think that end users are going to be more willing to ask questions that they weren't before because I'm already seeing it now, even in our testing. I'm already seeing people ask questions that I know they would not have called and asked. So it's going to help them do their job better and of course, make our life easier. That's a really interesting point, Eric. And that's definitely something that we see across kind of the Autopilot user base is that people are embarrassed sometimes, you know, they're not going to open a ticket and ask, hey, how do I send to something in, you know, in PowerPoint or, you know, how do I create a transition or what does this error mean? And what we've seen is that they, and I see it in ChatGPT, right? I wouldn't want anybody to see the kind of questions that I'm asking ChatGPT, right? That's the last thing that I would want to expose to anybody. But we're seeing something similar with Autopilot, right? And I think that the other point that you mentioned, which is interesting is knowledge base, right? I think that all of the IT leaders that I've ever talked to, you know, you ask them, do you have a knowledge base? Sometimes yes, sometimes no. But always the answer is we don't like maintaining it. Nobody likes working on it. It's kind of very, very burdensome. And I think that what's interesting is that when people see that one, you know, the AI can help build a knowledge base, but then there's this kind of a self-reinforcing loop there where, you know, if they maintain that knowledge base, people are going to ask us questions, it's going to be surface for them. That's definitely something interesting that we see. Eric, staying with you for a second, what was it like? What was the discussion inside LAMPS plus about bringing in a virtual IT technician? Was there concern about how end users are going to adopt it? You know, we know that you've got different type of employees. You've got people in different parts of the business. What was that discussion like? What were we privy to during the process? You were privy to quite a lot. I speak in my mind, you know that. But, you know, there were some internal discussions that maybe I didn't share. So, you know, I think the spectrum was wide on that, right? I mean, obviously my standpoint was I think we can improve efficiencies and I think it's going to be a great experience. I also think that there's going to be a couple of holdouts that are, I do not want to talk to a bot. And that just is what it is. Certain people, it's the same with talking to a regular person. Some people like certain technicians and they want to talk to that. So, you know, I was aware of that. Now, the sentiment from leadership, I think, was, you know, prove it, show me, which we were able to do, right? Because you guys gave us some early access to be able to kind of show some flows. And I think it was one of those things where it was a 50-50, honestly. They were kind of like, yes or no. And they said, you know, let's just try it. Let's see what happens. The worst case, we can stop using it. Right. But I think overall, especially the way, I think it was all in how you sell it to leadership, because, you know, right now there's a lot of, I'll say like snake oil AI, so to speak, or like sort of like, I guess the best way to say is like workflow theater, like workflow stuff. That's not really AI, but it's like, oh, it automates this one little thing, you know? And there's a lot of everything. Complicated decision trees, right? If then statements. Exactly. And they're selling it as agentic AI, you know, and it's not right. So really getting into to it with leadership about true agentic AI and how it's not. And also Tara was doing this a little pre the big boom or at least starting to do it. You know, I think also helped, but, you know, the fact that there's actually an agent running on a machine that has pre-trained it-based, you know, information and that there's multiple LLMs to kind of. You know, shift and say what to do and that we have some control over it. I think very quickly it started to catch on with the leadership and they were starting to think, hmm, you know, we are cutting staff and maybe this is a way to augment our existing staff. So I think, like I said, it was a 50-50 with who you talk to, but for the most part, we obviously tipped the scales in the right direction and we got to go. Awesome. America, kind of a question for you. I think one of the biggest concerns that we hear now in this new space of AI and agentic AI is security. Security, privacy. Can you talk a little bit about what that looked like inside your organization? How much kind of did you have to do while convincing the organization that it's time to adopt AI? How much of an appetite was there kind of prior to you bringing Autopilot into the organization? Great question. From a security perspective, I mean, we evaluate like every piece of software that we use, every bit of technology that we use, like based on that individual company, and so, I mean, there are some platforms that we've opted to not use because of like where data is stored or how it might be accessed and just generally how models are trained. So you all beat that test for us. In terms of like the appetite to adopt AI, it's like a really big buzzword in our industry specifically right now. A lot of other, like just even property management software companies are not just selling like their own AI agents within their tech stack, but also starting to speak more to agentic AI and how that's kind of going to shift things in the industry. And for us, as more of like a small midsize management company, we were really looking for where can we adopt this internally first in a really low risk way so that we can kind of understand like what's really needed, how involved do we need to be in order to adopt this kind of technology like really strategically. And so it ended up just kind of working out that you all offered a piece of tech or a solution for us really in that way where we could expose it. We did first expose it to our corporate end users with the nuance that like, Hey, we're still working on the knowledge base behind this to make sure that it's really, that it really understands you, who you are, what you use in your use case some of the more frequent issues that you see, but for general IT issues it should be like 90% of the way there at minimum. Please give us your feedback. Here's how to communicate with it. That's also been a unique challenge and something that we discussed at the leadership level is how do we make sure that our company is really prepared to engage with the agent? And so just a little bit of like, like forward communication about prompt writing, I think came in handy there as well. That's an interesting, that's an interesting one because I think that, what we talk about a lot inside Atera is that Autopilot, Atera as a business was a B2B business for all these years and Autopilot all of a sudden is working with end users, right? And I think that that education for the end users and telling them what it is. And you just mentioned in America about, you started with the corporate level, but what did that communication plan look like? How did you kind of communicate with the internal end users about this product? What did you find to be effective in helping people over that, maybe initial hesitance? Yeah, baby steps for sure. A lot of communication to our end users. We primarily are a Microsoft shop, so a lot of our communication comes out as like a news post, for example, where everyone can kind of see that information centrally. And then of course there's the email route. But a lot of like baby steps really like, hey, like as an IT department, we've opted into using Atera. Here's why we decided to use it. Here's why we think it's so great. And here's kind of like future looking, this AI component that we are going to adopt soon that we're really excited to show you about. And like, here's how it benefits you. You're going to get faster responses. You're going to get some initial troubleshooting steps before, like maybe before it gets escalated, before someone can talk to you one-on-one, schedule a call with you, work around your schedule kind of thing. So we really tried to sell it to them upfront that way. And then as we got closer and closer to going live with our corporate group, here are some tips on how to get the most out of it, right? Here's how to structure your prompts, like, or here's what a prompt is to begin with. Here's how to structure a prompt, like give some detail about what kind of issue you're having, maybe steps to recreate the issue that you're having, maybe a bit more about like you and your specific use case just while we're building on that kind of information for the agent itself. You won't always have to provide this amount of information, but just a good practice to really fully write your prompt and structure them before you even submit a ticket. So as an IT leader, what's your advice for other IT leaders out there, people that are considering autopilot, considering AI kind of, what's your advice for them? How do they get their organization on board? You know, I was listening to America talking. I'm like, I want her advice on how she communicated. We're about to ask her for her advice. I think you have to start first, Eric. Yeah. Yeah. Oh, you know, but, um, I mean, I think, look, at the end of the day, security should always come first. I think that's huge. Um, my advice for IT leadership, pick partners carefully. Um, one thing that's important to, to me, um, is, you know, and a big reason why I really pushed for a Tara and I think, and we got it through the door is because a Tara has like collaboration in their DNA. They're, they're very like open to, you know, it, it doesn't take much for me to get a leader at a Tara, um, on a call. Um, it's happened already multiple times. So, you know, um, that, that sense of collaboration and teamwork, um, I think is important, um, when, when picking a partner, um, and, you know, scope, scope, what you want to do, you know, I mean, the sky's the limit, you'll understand what you actually want to accomplish. Um, I think that, um, you know, we're in a world now where a lot of teams are understaffed, you know, that's just, that's just a truthfully matter. Um, and you know, if you have a secure tool that can help improve your efficiencies and help, you know, even resolve, you know, let's say a quarter of level one tickets, you know, I'm actually finding too, as I make the system more and more and really get into it, that it's not just solving those level one, the level one stuff, it's actually also preventing escalations that would have come up to tier two and tier three that get stopped by either the technician or the user because of the qualifying questions and the autopilot being involved. So it's actually not just solving those level one things. It's actually, you know, stopping some escalations that don't need to happen. So thinking about those kinds of things and thinking, kind of playing the tape through of what these tools can do, um, to better plan your scope and, and choose a partner. Sage advice, uh, America, I want to give you kind of the final word here. And, and I'd love kind of to hear what would your guidance be to it leaders, to it teams, people that want to bring. Agentic AI and AI agents generally into their workflows. What's your advice? How would you advise them to think about that kind of project and process? Um, to start with, I would say, um, I mean, I feel most organizations have already made the decision to, to try to find a use case for, for AI internally. Anyway, I, our perspective is that starting, um, internally with your internal customers is going to be your best bet, so it was a pretty easy decision to start with the it team. Um, or, or a similar, um, internal facing team. Um, Step number one is to put your, your end users first. Once you've decided to adopt an AI agent of sorts, um, being really proactive and preparing them for adopting AI, uh, communicating requirements and communicating, um, you know, what it's designed to do, why it's worth trying, um, really, uh, bridging that gap is essential. Um, to Eric's point, we're, we're tech technical leaders for technology leaders. Um, we're really used to hearing about AI. It feels familiar. It feels intuitive for us, but for many end users, it's really still a new concept. Um, we found, uh, the welcome tour for a terrorist specifically, very helpful. So adopting something along those lines to again, sell it, provide the end user a clear understanding of what types of issues the agent can help with, um, help build trust and encourage them to kind of use it from day one. Um, and then also investing time in any of your internal knowledge base or any documentation that you use for any agent really is important, um, for us in particular, structuring the articles to help the, the agent identify when context is relevant, um, and then also guiding users to provide the right contact. Content by providing examples of questions that you might want to ask to, to gather the context necessary to resolve an issue. Um, walking through diagnostics and troubleshooting steps in that knowledge base is also really critical. Um, so the more tailored and intuitive your knowledge base is, the more capable and accurate the AI becomes. Awesome. It sounds like you put a lot of emphasis in that answer kind of on the end users America. So it sounds to me like, do you see the main beneficiary of autopilot being those end users, which are getting quicker resolutions, quicker responses, rather than the technicians, which don't need to deal with those problems any longer? The happier your end users are, the happier your technicians will be. Because by the time an issue comes to them, uh, the end user is still, still happy for the most part. Um, I, like I said earlier, like the sentiment has improved just, just generally, um, something about emojis, man, I don't know what it is. Something about emojis. That's awesome. Um, America, Eric, I can't thank you guys enough for all of the time that you dedicated to us and for being a wonderful customers of Eterra and sharing your wisdom kind of with the crowd. Thank you so much for being here. Um, and, uh, thank you to everyone who's tuned in and listened. Feel free to reach out with any questions you might have.
