Making AI Customer Support More Human with Asaf Goldstein

Russ Fordyce 🏆 (00:01.094) Welcome back to the Winner's Circle. Today we're here with Asaf from SysAid. They are on the cusp of customer service and AI and trying to make AI a little bit more human. They are a recent Excellence in Customer Service award winner. So Asaf, welcome to the Winner's Circle and congratulations on the award. Asaf Goldstein (00:19.554) Thank you very much, Russ. I'm happy to be here and thank you for inviting me. Russ Fordyce 🏆 (00:23.204) Absolutely. Well, give us a little bit of an elevator pitch on Cis-8 and what you guys are doing. Asaf Goldstein (00:28.014) Sure, C-SED has been around for the last 25 years. We're having a SaaS software for ITSM companies. We're basically providing all-in-one ticketing, asset management, license management, and obviously AI tool for IT departments, but not only. And with our software, IT admins can basically manage everything when it comes to IT using our tool and with the help of AI even enhance and proactively work with their end users. Russ Fordyce 🏆 (01:10.128) Yeah, now everybody's talking about AI and customer support. I think it's probably one of those natural areas where, you know, it can probably do 80 to 90 % of a job that a human can do. Maybe there's a little bit of gray area there. From where you sit, what actually is changing right now? Asaf Goldstein (01:26.776) So I think the most important thing that support leaders need to understand is that support needs to change from being reactive to being proactive and to drive even revenue and not just be a cost center. think that with the right amount of human in the loop with the collaboration of AI, we can definitely become more proactive and we can identify issues before customers even notice them. And perform deeper analysis, perform and have a better collaboration and connection with our customers. Because at the end, the way that I see it, it all comes down to the customer experience. It doesn't matter how good your product is. If you don't have a good team that supports it, not just customer care, but also customer success, relevant product and so on. I think that you're not going to be a successful company. And I think it comes down to. how we can use the technology with the relevant people to actually make, to show value to our customers. Russ Fordyce 🏆 (02:33.744) Now you're responsible for global customer care. What does that look like in an AI driven environment today? Asaf Goldstein (02:39.66) Yeah, so I manage a department of around 30 individuals in different tiers from tier one, tier two, tier three, and the technical account managers. And basically, arrive when customers have issues, when something is broken and they're frustrated and we're trying to save the day. But when AI comes into the picture, We can identify trends. can identify things that happened before to show them before the customer even experiences them. There were a lot of cases that we, with the help of AI, identified an issue before the customer even saw it. We told them, look, we identified something. We're already on it. Don't worry. And within a few minutes, it was resolved. Another good example is that we started a new community using Discord. Once two customers or three customers said that they had a critical issue, our AI automatically identified it, checked our internal tools, saw that we have an issue. In five minutes, we had everyone from engineering and DevOps already in a war room. And in 15 minutes, the critical issue was resolved. And that's exactly where we play around with AI and people to make sure that things are being delivered faster. And the customers were actually happy. We had a critical issue. They were down for 15 minutes, but they said like, wow, okay, I'm happy. was solved so quickly. Russ Fordyce 🏆 (04:22.13) Well, it's funny, I was in telecom for a long time and there was a theory, I don't know if anybody ever put it into practice, that if you had a customer call into customer service and have a good experience, they were more loyal than one that hadn't, that had just had perfect service all those years. So it's interesting that you say that because I think people are naturally, if you help them out, they realize things go wrong and things are gonna happen. But if you do it in a kind of, say, in a very reactive, proactive method, they're going to be really happy with that. I imagine that increases loyalty quite a bit. Asaf Goldstein (04:57.708) Definitely. So I think one of, one of our customers said something that really, you know, stuck with me. He said, it's the first time that your support department gave me a, that the support department in general gave me a wow experience. And that's exactly, that's a quote that I'm using with my entire department. And I'm telling them, we need to replicate this. need to duplicate that experience because if a customer is telling us. You gave me a wow experience. He'll definitely be more loyal to us. We'll reduce churn. We'll increase the, the adoption and stickiness. And, and a lot of it comes down to a, value that we're giving them as support as well. Russ Fordyce 🏆 (05:41.36) Yeah, now with results like that, well, first of all, give that information to your marketing team too as a marketer. I want that quote and that customer's name and blast that all over the place. There seems to be a lot of pressure to automate support. Where do you think most of that pressure comes from today? Asaf Goldstein (05:57.144) So I think when it comes to automating support, I think there's a lot of things that at least the basic how tos, the basic simple questions can be answered with the help of AI and the co-pilot and so on. One of the things that we did when we just started implementing AI, we actually assigned a dedicated project manager. Her job was to review all of the interactions to make sure the answers are valid, to validate them, to add more and more information to our data pool. And we saw the results because over around 12 % on a monthly basis of all of our tickets are being resolved completely by AI. This means that my agents can work on actual complicated tickets on the things that required the human touch, the things that require a remote session or something that is more complicated. And we leave the more basic stuff to AI. So they handle less tickets, but the tickets that they actually handle are more complex, require the human touch. They also get information from AI that actually tried to solve it before. So AI tells them, I already tried ABC with the customer. You don't need to do it. You can start from the more advanced troubleshooting. So it also saves time and we can focus on the more complex scenarios and solve them faster. Russ Fordyce 🏆 (07:30.47) Yeah, some people don't realize that there's tiered support in almost every organization, you know, and more technically you get the fewer people at the top, right? And those are those brains that you absolutely have to have because they're solving complex problems. I wonder, you know, I wonder as we get into this, is there a risk of over automating all these customer interactions? Asaf Goldstein (07:50.84) So definitely, I think there is a risk. Like you don't want to get to a point where customers are frustrated with AI and frustrated with the automation. So you need to find the right balance that when we see that we can't help a customer using AI, it will direct them to an agent and it will tell the customer, Hey, look, I already gave the agent all of the relevant information. can tell you from a personal experience that I had with some airline company. that I tried to get just a simple invoice sent to me. And the AI kept telling me, yeah, I can do it for you. I can do it for you. And every time I tried, it just didn't work. And he didn't want to give me a live agent. took me like, I think 30 minutes just to ask him, okay, you're not helping me. I need a live agent. And I think that's exactly when you're over trying with AI instead of understanding, okay. This is where I need to stop and I need to move it to a real person hand and have the human in the loop. Russ Fordyce 🏆 (08:55.154) Well, it's so interesting. In technology, we always talk about stacks or layers. You might have your technology stack of what are the platforms and tools that you're using. We have a stack in the transport world going from fiber all the way up to the application layer. And now we're really talking about a new layer. It's always been there, but it's always been a different type of interaction, this human layer. How do you define that human layer in the modern subculture? organization. Asaf Goldstein (09:26.424) So I think nowadays, especially with technology all around us and all of the vibe coding apps and everything that exists, there is one thing that AI and all of those tools can never replace and that's the human personal touch. There are countless of examples that we had where customers asked for a remote session or ask someone to actually jump on a call with them, even travel abroad and be with the customer to help them solve a complex issue. Those things can, I think, can never be replaced with AI and all of the tools and with all of the technology that exists around the world. I think that's something that definitely will stick around for a long time. Russ Fordyce 🏆 (10:18.332) Well, now you said the company's been in business for 20 years. You've got kind of all this historical data and expertise. What are those moments when humans are still absolutely critical? Asaf Goldstein (10:30.902) so definitely, think that, I'll give one example, when, when customers are not happy, it, always happened. can't avoid it entirely, but we can minimize it. And, before I joined, I joined like a few years ago, we got to a point where we had on a quarterly basis around the 40 unhappy customers out of, let's say 300 that left surveys. That's a huge number of unsatisfied customers. And I'm just finishing, you know, the Q1 QBR we got in this quarter to five. So the decrease from 40 to five is amazing. And that exactly where our changes and the people that are critical make an impact. Because for example, every unsatisfied customer Russ Fordyce 🏆 (11:10.481) Wow. Asaf Goldstein (11:26.05) gets a personal call and email from one of the team leads, not from an agent, not from someone like a team lead that is responsible for that agent tells him, I want to get on a call with you. And I want to understand what we could have done better, what we could have improved. We validate with the customer a few times that everything is solved, that everything is working as it should before we actually close the ticket. We're doing a lot of things that are critical in terms of human and AI doesn't, they don't understand those. small sentiment when it comes to, okay, I'll just close the case and I won't ask if there is anything else. want validated there, there aren't any other issues and I won't close the loop when it comes to unsatisfied customers. And I can say that even when we have an unsatisfied customer from a, an AI response, a team lead also calls them and say, okay, why the AI didn't give you the right response. How can we improve it? So next time you will enjoy a better experience overall, because it's frustrating. It's frustrating. doesn't matter if you, if you're a contacting your cell phone company or you're contacting your local authority or you're contacting a SaaS business, you want a good experience. And the good experience comes from the combination of AI with people. And again, that's, think, how you can create loyalty, how you can create trust. Russ Fordyce 🏆 (13:00.272) Yeah, well, it's interesting. The empathy is kind of, I think what you're driving at is you need that human empathy. When something's gone wrong, you need that human empathy to kind of reach out and say, hey, I hear you. I hear you're struggling. Maybe we're the ones causing your struggle or whatever. But just having that human touch, I think really matters. So if we think we're going to work all the humans out of jobs, think again, because human empathy is really, really hard to emulate, I think. So you've been using AI in this kind of blended model for a bit now. Where does AI work really well and where does it fall short? Asaf Goldstein (13:36.28) So I think I'll give a few examples. think that from my perspective as the director of the department, it really helps me understand bigger trends across thousands of customers and understand, okay, where do I need to take it with product and R &D because we have a frustration in those areas of the product. And I can definitely leverage that information before AI I needed to do all the manual work and understand Excel's and like look into data and try to understand what I want. And with AI, I'm doing it super fast. can get a lot of valuable information. And also on the agent level, we have AI agents that we created that analyzes the sentiment of the customer, not only now, but also from their previous ticket. So if we had a customer that had a lot of issues and was frustrated, the agent on the tier one level automatically gets an AI agent telling them, hey, look, this customer opened 20 tickets in the last year, and he was very upset that they took a lot of time to resolve. So let's try and do something different this time. And it gives them a list of relevant recommendations and action items on how to approach this case in a better and a different way. When it comes short, I think that a lot of the times we still see hallucinations in all of the AI tools that gives you like, okay, yeah, I analyzed and this is what I come up with. And when you actually digest the data and investigate it, you see that it's not actually a hundred percent right. So I think that hallucinations, I see that a lot as a critical. issue when it comes to delivering better service and tools overall. think that that's something that needs to be solved. Russ Fordyce 🏆 (15:38.096) Yeah. Now the balance here is really important between kind of that human and the automated or the AI. How do you decide kind of where your process goes into automated or into human? Asaf Goldstein (15:53.016) So generally speaking, we try to automate everything that we can on the more strategic areas and to solve things that affect a lot of our customers. I can say that all of our customers go through a basic answer from Co-Pilot. If it helped you, good. If not, you can... decide to continue a bit more of it, or we can already direct you to a live agent. And I think a lot of the things that we're doing are actually internal. The customers don't see them because the agent can automatically see everything they need to know from the AI agents that actually digest and analyze everything there is about this customer. So. even which platforms they're using in their ecosystem and what is the customer description and job title and so on. And we give them and we give the agent that analysis so he can provide more of that empathy that you mentioned and the value to the customer. And I think that's exactly where we play around with AI and the human. Russ Fordyce 🏆 (17:14.502) Now, I think everybody's worried that AI is going to come for all of our jobs, right? And we talked about kind of the tiered level support. In your world, how has AI changed the role of your support agents? Asaf Goldstein (17:28.152) So definitely I think that it's a matter of let's start with the simple thing. think it a matter of how do you use the tools and how do you use it to help your job and not replace you. I think that when it comes down to the more basic troubleshooting or how-tos, I would say AI in the next few years could definitely replace people that are doing the basic stuff. So my suggestion to all of my agents is like, okay, see how you can progress further. Because maybe in two years, I won't need tier one anymore. I will need only tier two and three. I will need SMEs that are experts in different areas that potentially AI it will take time for him for it to replace you. But definitely I think that if you don't know, and you don't have the ability to understand how AI can help you, then you're you're in danger of being replaced. And definitely I would say, you know, always learn always try to see what what you can improve yourself how you can use those tools to your benefit. I don't remember life before utilizing AI. And I can say that it definitely helped me get to those results and not replace me. Russ Fordyce 🏆 (19:04.71) Well, it's so true because this is a fundamental shift in society and technology and it is changing everything that we do. I think those task-driven roles are going to be automated and taken over and that's what people are really worried about. As you evolve in these customer care teams, what skills do you see that are becoming more important than ever for humans? Asaf Goldstein (19:25.454) So I think that anything that would revolve around deep technical skills that require specific knowledge in specific platforms or something that will take time to actually understand and analyze, that's something that you need to focus on. At the end, think that there are a lot of things that AI can definitely replace, but there are things where AI can just assist you instead. I think, you know, at some point, I would say like a few years for now, we would stay, we would only have people that will manage AI instead of managing customers. So people that will... enhance the AI with more connections and more information and enhance the data pool and the coverage. And we will see more jobs like AI manager rather than customer care manager. And I think that's where the world is going. Russ Fordyce 🏆 (20:38.908) Yeah, I hadn't thought about that. I think, I don't know if I was the first, but I was coining that 2026 was the year of the CAIO or the executive, the chief AI officer. But I think you've got a great point, which is you're almost a product manager inside of a customer support organization, kind of managing that AI experience, which is really interesting. And you mentioned that wow experience your customer had. What do you think makes kind of one of those interactions feel good or while, you know, when AI is involved. Asaf Goldstein (21:13.742) So I think that when we come to a customer, doesn't matter the size or how long they're with us. And with the help of AI, we're already coming to them with all of the information upfront. And we tell them, hey, Ross, know that you're the chief recognition officer in the business intelligence group. I know that you already opened five tickets with us about XYZ. And I know that you're using play. You're using Azure and Slack and whatever. And I already have a good place where I start with you and you already feel confident that, okay, I know what I'm talking about. I know who you are. You don't need to explain yourself to me. And I think that's exactly where we see AI helping us. Get her more information about customers and, and giving us. recommendations and analysis specifically on each customer. And that actually creates this wow experience because okay, you know me, you know what I need, you know what I had in the past that creates the experience that we're looking for. Russ Fordyce 🏆 (22:27.202) Do you think that the kind of your customer expectations are going to change as AI becomes more common? Asaf Goldstein (22:33.774) I definitely think that they would expect a even higher standards from us. And that's exactly my role to, to understand what they're going to expect from us and how I can keep on replicating this wow experience when it comes to a world that is evolving around AI. And, I think that's why I love my challenge because it involves technology and people. And that's why I chose this career path in general. I always try to think about, you know, what's going to be in 2027, what's going to be in 2028. How can I create this experience and make sure that our customers keep on a loving us and drive loyalty and drive more adoption. And that's exactly what I aim for. Russ Fordyce 🏆 (23:27.142) So you've done this now, you've rolled out AI in your organization, and I'm sure you ran into a couple of mistakes along the way and fumbles. And even in your customer's perspective, what are you seeing out there where customers are kind of falling into a pit or making a mistake when they're rolling out AI into the support channel? Asaf Goldstein (23:47.352) So I would say that the first thing that we did, as we said, as I said, is assigning a project manager that was responsible of validating answers. think that if you don't pave the way in the beginning, AI is going to make a lot of mistakes because he, he, in the beginning, it just tries based on the data that you feed him. But if you don't have someone that actually. a performs the relevant guardrails and validates the data pool answers and even formats it in the way that your organization sounds because someone is more formal. Someone is more, you know, going easy with the answers. And I think it's super important that you will have this person that actually validates everything and making sure that customers are, that the AI answers according to what their customers need because I think that if a customer will get the wrong answer from AI in the beginning, he will just say, okay, I'm not going to play with it again. I want a person immediately. But if they will get the right answer, they will say, okay, I'm going to try it again. I'm going to try it again. And if I'll need a person, then I'll get to it. And that's in my experience and my personal approach as well. Like I prefer that AI will solve 80 % of my issues. But for the 20 % that I actually need a person that he will know to like as fast as possible to direct me to a person. Russ Fordyce 🏆 (25:23.302) Yeah, we went through a similar transition 20 years ago when all of support was on the phone. And then we went to email and then chat. people thought that, nobody's going to go to chat for help. Who wants to sit there on a computer and ask for your bill or for anything else? When in reality, we'd much rather talk to a chat agent than have to be on the phone talking to a human with language barriers, time delays, whatever the complexities were. I imagine it's almost like that of, we've got to get used to it a little bit and get a little more comfortable with it. That AI manager you're talking about, what sort of metrics are they looking at kind of on a daily, weekly basis that kind of direct them? Asaf Goldstein (26:10.03) So there are few things. first thing is the quality score. So in our system, every answer has a quality score that is based on the data pool, based on previous answers and a lot of different metrics that we're taking into account. So Hergold was to improve the overall quality score of the answers. So for example, if in the beginning it was Like 50 % because I found something that sounds similar in one of the articles. Her goal was to say, okay, you need to look at a article ABC to improve it to get to 90 or a hundred percent in the quality score. And the second thing, and probably the most important one is the tickets deflected. So if we had a hundred tickets in the past and now all of them go through AI, we want that at least 12 % of them would be solved by AI. And we want to increase this number as we, as we progress. Um, I think those were, and the other metrics are not directly related to AI, but they are a side effect of improving this because if in the past I had a hundred tickets that came to my agents and each one of them took an hour to get to now I only have ADA tickets because AI solved 12%. Then I would expect my first response time. to be faster because I'm handling less tickets or, you know, solving things faster, improving customer satisfaction because I have less tickets. All of those metrics correlate directly to the success of the AI implementation. Russ Fordyce 🏆 (27:52.819) Now, you, like I said before, you've been at this for a little bit in this kind of role, and also the company's been in this customer service kind of IT support business for a long time. If we're talking to another customer support manager or somebody that's looking to kind of AIFI their organization, what principles do you think they should remember to kind of not lose sight of? Asaf Goldstein (28:15.746) So I think first of all, one thing that was super important for us were guardrails, right? We don't want AI to give answers about different customers. That's something that was super important for us, like to understand, okay, what are your limits, what you can do, what you can do. think that also when it comes to making sure that you keep on validating the AI, don't just implement it and say, okay, you're good to go. You constantly need to monitor it and make sure that it's working as you expect. And I think at the end, just keep on identifying more things that AI can do for you. I'll just give quick example. I talked to an AI agent of like my water company and they needed an issue. needed the help. I didn't talk to a person. told them, I told the AI what was my issue. He helped me coordinate the technician to my house. We using the AI. didn't need to go through any, like it was an amazing experience for me to get a technician to come to my house to fix my, my tap. So I think that like, that's, that's the experience that the, the, managers need to think about, like, how can you make Russ Fordyce 🏆 (29:34.257) Yeah. Russ Fordyce 🏆 (29:43.92) Wow them. How can you wow them? Yeah. Awesome. Thank you so much for joining and congratulations again on the excellence in customer service. The judges were very clear that your solution is out there really making a positive impact to all these IT organizations. So congratulations again and thanks for joining. Asaf Goldstein (29:44.972) Yeah, exactly. Asaf Goldstein (30:01.826) Thank you very much. Thank you for having me.