AI Readiness for Customer Service Teams

Hello, everyone. Good afternoon or good morning, depending on where however you are. Today, I will be talking about AI readiness for customer service. In twenty twenty six, AI readiness is the word that we are all looking for. A year and a half ago, I was at Microsoft's headquarter in Redmond at a conference where one of the tech evangelists talked about how AI is going to shape the future, and it's not going to be in the the sense that we all think about. He actually compared it to to the smartphone evolution. One of the biggest change contrary to to how many very many people think in the smartphone revolution was actually the GPS, the the inclusion of the GPS within the phone itself. And that enables so many things, like navigation and and being able to use the apps, being able to search on Google Maps, find restaurants near you. And that was one of the hidden powers of of the smartphone, and that had an outsized impact on on how it was was used. And he liked and likened the GPS solution within the smartphones to to how AI will be used in customer service. They anticipate that that will be the area with the biggest impact. And not only because we are we are talking about everybody's when we say customer service or AI in customer service, everybody starts thinking about chatbots picking up the phone and just chatting with these very clunky experiences. And that's that's not what we what we want here. What we what we really want and what we talk about is is getting efficiency and and gaining momentum within the customer service business and improving customer experience using agentic technology. And that's exactly what I will show you today. Because the ones who win the AI race in in twenty twenty six are the ones who are taking small steps forward and the ones who experimenting while getting value out of AI. It's very much dependent on where you are in your journey. And at every single stage and every single step, you can start making good value out of AI with AI. So with that, allow me to introduce myself. I am Balaj Svarovat. I'm the founder of Visual Labs, and we have built the company into one of the one of the prime Hungary based Microsoft partners. This is our wall of pride. All our colleagues certificates that we that we hold and have in in the company and, obviously, our solution designations, which are the badges that we receive from from Microsoft. And and what's really at stake and what what's really important here is the companies who embrace AI readiness will be able to build agents two point five times faster than the companies who don't. And and that speed, that learning scales up and improves over time. So we want to make sure that that you can get started as early as possible because the earlier you get started, the more benefits you'll you'll start reaping. And what we see when we speak with our clients is some of the misconceptions is some people say, Oh, AI is easy. You can just chat with a chatbot and you're already using AI. But in reality, to get proper AI transformation and get proper AI value, you need to have organizational readiness, and you need to get those readiness pieces sorted first. Quite on the other hand, very many people think that AI must be perfect. Everything must be perfect before we get going. But in reality, if you start get going and just have a clean path of, you know, where you're stepping, then AI value will will be realized without having to have the whole room in order. And another one, probably the most common misconception about artificial intelligence in in today's world is Copilot is AI transformation. But in reality, agents will bring proper business value. Copilot just bring efficiency improvement. So that's why we believe it's so important that people focus on on the foundations based on which agents can be built and and can can be improved. So before we get going, I'd like I want to make sure that we that we clarify some of the often heard terminologies, and this is this is something that we do at every one of every episode. So I wanna make sure that we cover what Microsoft calls a frontier firm, which is a human led but AI operated company. This is where AI agents work alongside people to run some of the core business processes. Let me talk about why this is so important. Because because what we find found here is in in legacy organizations in in today's world, we have capacity gaps, knowledge gaps, and a lot of slow and manual processes that need to be attended to with tedious tedious work with very, very varying quality. And nowhere this is more apparent than in customer service. We've got troughs and peaks of resource demands. So we have people who've been doing in the same position, and they know everything in and out, and we have lots of lots of people who are churning. And therefore, there are massive knowledge gaps and maintaining the knowledge article knowledge base is, is more than a full time job. And there's so much admin that needs to be taken, especially in highly governed areas such as insurance. And that results in very varying quality, how people actually attend to those tasks. And in today's world, everything is getting more expensive with inflation, with rising labor costs. That's that's we are almost at an unmaintainable level. But with the Frontier firm, which is a human led AI operated company, we've got scalable digital workforce who help us automate things and and scale up based on the demand. We've got an enterprise knowledge engine, which removes the human bottlenecks, and we've got automated workflows and processes that work outside of a regular working hour. And, of course, we are not we are not just getting rid of humans and the human touch of an organization. We're talking about the human agent collaboration where we can remove the low level tedious tasks and just focus on high quality creative labor. And this will result in faster innovation, faster time to market, and better customer experience ultimately. So just getting back to sorting the socks. Next up is our Copilot, which is already mentioned, the AI assistant who is designed to help improve the human workers' efficiency. And this is also Microsoft brand for AI, the UI for AI. And now we are arriving at agents. Agents are the AI assistants designed to automate and execute business processes working with or working for a human. And let's dive a little bit deeper into that. So what are agents? Agents are AI assistants designed to automate or execute business processes. I just mentioned that. And this is how an agent is built up according to Microsoft. It has an orchestrator. The orchestrator decides what to pull on, what skill to attend to. It's got user experience. It's got knowledge, skills, and autonomy. And it has some it's of course built on a foundation model and agents can be linked together to have a multi agent architecture, which is increasingly what we're seeing. And there are spectrum of agents. Are even today's chatbots can be considered agents because they have skills they can can they can attend to. They can create a Word document or or an Excel sheet nowadays. So those can be considered agents. You can upload a PDF or or any sort of file, and they would understand this. So those are the retrieval agents. They can undertake basic tasks, or they can they can work in a fully autonomous manner. And that's that's what we that's what we mean when when we talk about the the spectrum of agents. And Microsoft has identified customer service scenarios that arrive for for agents, And they have this quadrant, the baseball card quadrant. The ones that are low effort and high value initiatives, they call those the home runs. They're the high effort, high value initiatives, which are strategic imperatives, something that you want to embark on, but you know that it's going to take a lot of work to get there. So one of the one of the areas that they they recommend looking at is document compliance agent, right, which is which is just just at that stage. That's my highlighter there with document compliance. There's so much documentation that needs to be done in customer services. You want to make sure that those documents are hitting a certain standard, and you can you can use all sorts of compliance metrics as soon as you have those those things defined. You've got the knowledge metadata agents, all the knowledge articles, being able to find them and being able to tag those those those articles is is a job and a half in itself. So being able to utilize an agent to just tag and categorize in all your articles that are already existing is is definitely an an agentic use case where AI can be used very efficiently. And then there are the the areas such as the customer satisfaction prediction and the personalized customer upsell cross sell agent, where customer service representatives are involved in up selling and cross selling and being able to identify based on customer behaviors what we could do. So it's all well and good that Microsoft have put these together. But how do we do we actually get started? How can we start looking at the different areas within within our customer service, and how can we start reaping the benefits of AI? This is where I came up with with this framework for customer services. I've seen large customer service centers with with hundreds of call center workers who've been taking calls at three shifts a day, and they had two two big screens in front of them, about eight different windows, and they were just clicking around and and and navigating their way forward. It was it was just insane looking at the the amount of context switching they had to do. And that was a fairly automated matter. But we also have lots of small companies, customer service function. Sometimes there isn't even a separate support or customer service team. Someone just takes a phone call and would respond to to customer requests, or they would just fly in via email, and they would just do it out of a mailbox. So that's that's the first step where everything is manual. Nothing is automated. Everything is fairly ad hoc. It's just based on who's who's getting to it faster. Quite often, there would be a shared mailbox, and everybody's picking up emails on a first come first serve basis. And the next level from there is moving into a more managed customer service where we are actually categorizing the tickets. We already have a ticketing system, we are logging the items that are coming in, logging the responses. We we are able to track the the time spent, the number of tickets that have come in. And as we move forward, we can start introducing automation. Tickets are logged automatically. They are closed automatically. Basic responses are sent out to the client as notifications. And and at the top of the top of the list is the AI operated customer service, which is quite frankly not something that I've seen, and this is where we are still just getting to, in my opinion. If you're considering using proper customer service and the AI enabled customer service function, you should be evaluating where you are currently and what you need to do to get to the next step. And because we are still a bit far away from realistically being able to do a proper AI operated customer service function. So if we look at what we need to do at each level to actually move on and graduate to the next step, if you look at just the manual piece, for us to move on to managed, we need to we need to look at implementing a proper ticketing system. Right? That's that's step one. Your AI baseline at the manual level is just chatting with Copilot, copy and pasting emails in, hey, here's my email. This is what I got. Why should I respond to it? It's not getting the context that it needs, so you cannot get proper value from AI there. Your opportunity here that helps you move on to the next step is if you can automate your ticket creation and get ticket summaries based by using AgenTik AI. And that's exactly what I will show you in a few minutes, how you can just start using teams and email and a fairly simple flow to start automating ticket creation so that you can move up to the next next step on the ladder, which is the manage piece. So once you're at managed, if you want to move over onto manage, you want to get your cases, your issues, KPIs tracked. You want to build a knowledge base so that you have something you can improve and track against. And if you are not doing that, if you don't wanna move over or if you cannot just yet move over to the automated piece, your AI baseline will be just searching and categorizing using your existing ticketing system. You can use AI to retrieve information or categorize incoming tickets, which is still better than nothing. Right? So that's that's already AI value. But at this stage, you can start harnessing AI to introduce human in the loop AI flows, which I'll show you as well as part of the same demo, how you can how you can connect agentic AI to start taking off proper tedious tasks and a lot of copy and pasting from your human workers and outsource them to AI. Once you did your human in the loop and you've got AI flows in operation, that will move you to your automated piece. And that's where that's where we are talking about proper AI value with autonomous agents running end to end flows. And this is at at the automated stage stage, you're still not fully out operated, but you do have automated agents running in the background, making sure that you have human in the loop oversight. And to get you up to the next next level, you might want to consider introducing voice agents so that they pick up the phone instead of you. And you can you can create. If you're not moving forward, then you can start creating AI generated responses using AI. So that's that's another clever way of doing it. And and getting to that nirvana of an AI operated customer service center, you would have end to end case resolutions without human touch. And what you need to do, the AI opportunity there is to just fine tune it and keep evolving it and keep keep moving forward. Alright. So that was sort of the background. I wanted to give you the context of of the massive potential of AI that can have on customer service. We will eventually get to a stage where humans will be speaking, customers will be speaking to agentic AI over the voice and have proper end to end AI customer centers, customer service centers. But in order for us to get there, we need to move through every step of the ladder, implement the ticketing system, create a proper knowledge base, start using AI for searching, categorizing, and then automating processes. And and we cannot just leapfrog into Nirvana. At least that's what the current thinking does says. That's what we looked at. We we did our our research on this, and and all the big AI companies, technology companies, and advisors are saying this. With this, I would now like to move over to our demos, which have been readily prepared for you. We'll start with with a proper let me just make sure everything is all tied up. As you're most likely aware, Microsoft have their own customer service system within the Dynamics three sixty five stack. It's called Dynamics three sixty five customer service, and it has two aspects to it. One is the customer service hub, and the other one is the customer service workspace, which are basic ticketing systems. So in order for you to get to step one from manual to managed, you want to implement a ticketing system. There are so many systems out there. Some of them are even free. So I'll show you Dynamics three six five customer service as one option. And what you will see in a second, this is our customer service hub. You can see a a Power BI dashboard embedded here for demos purpose, and you can see the number of cases. So we're talking about cases and case resolutions when we are here, and this is where you will have all your customers, contacts, cases that are created. This is a fairly straightforward way of of doing things. Here's here's an active case that I can pick up. And the nice thing about this as as as part of the Microsoft ecosystem, there's nothing in this queue right now, but I can integrate this with Outlook, you know, the shared email emails and inboxes and have all the emails come directly in here so I no longer need to be working out of Outlook. So that's that's one way of doing customer customer service and and case management. And here's a more advanced way of of the same view. This is the customer service workspace, which allows customer service representatives to work in a single window without having to switch across apps. Notice you have the mobile phone icon here, which we can link up if you have a valid license. We have teams directly in here, and and we have old emails and inboxes. So if you have WhatsApp or emails, then we can just link it up here. And I have all these things available to me, what I need to be working on. And I have this clever specific case view that is already tracking SLA. You can see that I missed my first response KPI, but I have a resolve by KPI, which I can still hit, which is very, very clever. And I can have Copilot give me a highlight if there are there's enough information, and I can just send an email directly from here without having to jump out of the system. And I can use Copilot to to draft the email without me having to to put things in. And it already created a first version, and I can adjust it, make it friendlier, shorter, and let's go. And it's creating another response, and I can just send it directly from here or save it as a draft, and I can I can automatically set it for me to follow-up on on things? You can see that I have a draft email sitting here. And if I'm done, I can resolve the case directly from here. So to get to a stage where you've got where you have cases, all your incoming requests created as cases is is the step one. It's really important for you to to get rid of shared mailboxes. This gives you that auditability, traceability, and and it makes everyone's work more efficient because they don't need to be moving back and forth. And also, this will enable you to to start creating your own knowledge base. So knowledge doesn't sit in people's heads. They can just start creating. Here you can see my favorite knowledge base is about in Hungarian about payments and billing and coupons. So I can use that as a as a knowledge base here. So this is this is, again, step one. You can see AI already helps you as as it's built in here. But I think the real opportunity is is starting to automate these flows. And if I just show you if I just send an email from a client here. This is an actual email, a proper actual demo that you'll see. So what I have is I've got this customer complaint that they had an issue. Let me just share my my screen here. The rollout update Tuesday evening warehouse team hasn't been experiencing constant picking errors in Dynamics three sixty five warehouse management module. That's a real shame. Big lists are generating duplicate lines. Location directives seem to be ignoring our zone, severely impacting our daily dispatch, and the issue is in production environment, and the relevant entity is n t h u. This is from Test Alec from Novatec, Hungarian person. So I will just send this to our customer service email address and show you the flow that runs in the background that will be triggered right away. So here's my customer service issue creation flow that should will be running in a second as and when it gets picked up as it would be with demos. Just keep hitting refresh. As you can see, we tested this several times, We're waiting. If that doesn't load, what I'll do here is pick a just run a test instead of using the actual version. It might take a minute or two, there you go, for the email to come over, so we can see the flow running live. When there's a new email arriving, we will be running an AI prompt. That prompt ran for twelve seconds, and it's giving me a response in a very structured JSON format. And we are picking up that prompt, And that prompt will post into a Teams channel summarizing the issue and asking me for the next step. So we call this a human in the loop process because it's not actually taking an automatic action. It's asking me to verify if I'm happy with with the results, and the result has come in. Shall share the screen, and you'll be able to see what it picked up from that actual email. So here's my adaptive card created. Dynamics three sixty five warehouse management picking errors after Tuesday's update. And notice how it's actually not just copied and pasted over, but it actually summarized it. So this was auto generated, review and confirm before creating. So it's telling me that new case from email. And this is our customer. This is the company. This is the sender's email address, the references. So it's giving me details, and it's asking me it picked out the deadline, which was totally totally just free text in in the email as you saw. And it's giving me a summary Tuesday's evening exactly as we saw. And it's even recommending a next action to resolve, and I can I can edit before creating the case, or I can just click create case? And that will create the ticket in our ticketing system. So it's telling me thank you for my response. And let me just show you how it's actually running in the background and creating that ticket. Right here. Okay. So that has completed even though it keeps running and running. If I click it, what you can see here is it picked up and it created an issue in our own ticketing tool, which is Azure DevOps, which is an IT ticket management solution that we use. And it assigned it to me in the new status, and it picked up the exact same descriptions. So if I jump into Azure DevOps and hit refresh here, you will see that we have a new ticket created just now, and you can see the previous ticket created as well. Because this is AI, it's all used with generative AI. You can see the previous test with the exact same message that have a different title. This one didn't have D365, and this said after Tuesday update, and this is since Tuesday update. But if I open this up, it actually put in the issue description as it wanted, and it set the priority three, severity three, which are default values. But, obviously, I wanted to make sure that this is this is more urgent. I could set that as as well as part of the the mapping within the flow. So you can see how we went through this in real time in a matter of minutes. In reality, this would have taken several clicking back and forth thing, copy pasting into into the ticketing tool. And and this way, we just had AI pick up and summarize all the information and actually populate the fields all while keeping a human in the loop and make sure that they are involved. So this is this is where we are, and this is why we recommend making sure that you are using, taking every step. And this is why we're so focused on on readiness, that you are not, skipping these steps. You can move forward, using every one of these. So this is what I wanted to show you today, and let's talk a bit about some of the next steps and some of the things that we saw most often. We already talked about AI being easy. You can see that once I had that ticketing system in place, and I had my customers all sorted, putting together a five, six step flow with a clever little prompt wasn't that big of a deal. AI isn't that complicated, but making sure that we are covering all angles, we have a right ticketing system, we have the customers logged in our CRM system and it's connected to our ticketing system, that's the foundation work that needs to be done. Some people think that everything needs to be straightened out, getting things right first. But in reality, just need the foundations. Just let's make sure that you have a ticketing system, and then and then you can work around the rest. You can always put in a human in the loop to check for hours, making sure that there's nothing going array. We already have Copilot. Yeah. But will Copilot pick up your tickets and automatically create it for you? Well, that won't be a true transformation. You can only be chatting with Copilot. Right? And AI is expensive and it hallucinates. So you could see that I ran the exact same test and I actually used several email formats and different emails, and it didn't hallucinate once. It always got consistent same results. Today's AI models are so smart that they will just hallucinate if you are putting too much at it. They are so stable. So nowadays, hallucination is more of a misusing problem. Our data isn't ready. Plenty of our customers say that. And you just need to sort the critical bits, right? Who are the customers who are logging issues? Get that sorted. You'll be fine. Some of the objections are people won't adopt AI. Well, guess what? If it solves a true pain, if it saves them from copy and pasting, they will adopt it. They will keep on using it. And that's actually a big, big, big difference. And the other thing is they might not actually need to adopt AI because it just runs end to end. Right? So the real opportunity here is once you get on that agentic road and you start using AI, it will compound interest. So the better you start, the earlier you start using, the more return you will get as you move forward. How we help our clients is we have our own methodology, own framework, which is we identify the value first. What are the big levers that if you pull everything else will get sorted and will fall in its place? We look at your current systems architecture and make sure that what we propose and recommend fits in there. And based on that, we can build out an agent backlog. And only after this planning, we'll start getting into execution. And how we work is we have a one day AI readiness workshop, which is in person with all your decision makers in the room, and we will produce a ready to execute roadmap that you can just get on with. So if this is of interest to you, just scan that QR code or go to workshop dot visual labs dot com, and we'll be in touch and we'll I'll call you to see if we can help you and make sure that we can we can identify these these high lever areas, and we can workshop it out and plan your ninety day agentic road map. So thank you so much for joining me here for today. You saw how you can take it step by step and how at every maturity level you can utilize AI and you can utilize AI to help you move forward and the tremendous opportunity that sits within customer service. If this is of interest to you, if you want to get going, just do reach out, and we'll be happy to help you. Thank you very much.