AI Readiness for Field Service Teams

Alright. Good afternoon, everyone. We are just getting started. Welcome to the sixth episode of the AI Readiness series in quarter one of twenty twenty six, and this marks the halfway point of our series. Today, we will be covering AI Readiness for field service, and we have started the series, with an intro session in the first week of January that we talked about the different agentic technologies. Then we moved over to marketing, sales, customer service, and now we're talking about field service. So in today's session, I will show you what it means to be ready for AI in field service or field services areas. And this is this is one of the areas that that we work very heavily with and has been with me throughout my career. So in this series, we are covering quite a few things that that helps you helps you understand where where things are going with with AI and how to get ready for AI. Some of the common misconceptions that we that we see is, people think that AI is easy and, they can just turn on Copilot and get on with it. But in reality, finding proper value adding use cases is actually more complicated than, most people anticipate. So AI is actually easy once your organization is ready for it. Their misconception is people believe that AI must be perfect. Everything needs to be in order. Everything needs to be aligned to make value out of AI. But in reality, this is, this is we're still in the Wild West era of AI, so you better get going, start getting benefits out of it, and just see see how it works in a safe and secure way. So we recommend our our clients and and partners to just get going. And the other and probably the biggest misconception here is Copilot will yield AI transformations. But in reality, Copilot is not AI transformation. Copilot gives you slight benefits, efficiency gains, but it will not be transformative. It will not change how you do business. In reality, agents bring the value and not Copilot. So hello everyone. I am Barash Vorbath. I am the founder of Visual Labs, and, we've built the company from five people to thirty in, in about three years, and we are now three million in revenue with with twenty five great consultants, with all these badges, and we've got the necessary solution designations to be a proper accepted Microsoft partner. And we are in twenty twenty six, our team is doing lots of ERP and CRM and data projects, but we know that AI will be, at the forefront, of all the all the initiatives. We want to make sure that companies can get the right value out of their AI initiatives, and that's why we're focusing on the foundations. Because the companies who embrace Readiness will build build agents two point five times faster according to a Microsoft study, and that learning actually compounds over time, so people will, will start getting more value of it. And before we actually get into the subject matter of field service, I'd like to talk about some of the underlying definitions just very briefly. I can't stress these enough. So Microsoft has introduced the concept of a frontier firm, which is a human led but AI operated company where AI agents work alongside people to run the core business processes. And these frontier firms are emerging. They are among us. And there is Copilot, which is both a brand and a concept at the same time. Copilot is the user interface for AI in the Microsoft ecosystem, but it in its essence, it's an AI assistant designed to improve your workers' efficiency. And then we have agents. Agents are AI assistants designed to automate and execute the business process working with or for a human. So let me first talk about bit about why it's so important to start aiming for Frontier for why it's more than just a tech buzzword introduced by one of the largest tech companies. So what we are seeing is these legacy organizations that we are we are all living and we are part of, they face lots of challenges with capacity for human resources, knowledge gaps. There's lots of slow and manual processes. There are tedious tasks that need they're error prone, and they have varying quality. And and this, is done all amidst of rising operational costs resulting in slower, agility and slower innovation. And opposed to this, we've got the frontier firms and the opportunities that lie within the frontier firms where you have a scalable digital workforce, who are built on a enterprise knowledge engine. Therefore, you can remove the inconsistent, human execution and the human bottlenecks. You've got automated workflows and processes, that actually operate outside of regular working hours. You're not no longer constrained with work time and labor and and all of that. And you've got human agent collaboration, people working together. And once you've got this, you can start removing the low level tedious tasks and start getting consistent high tedious execution, which is why we want want to get into the frontier firm work land and essentially get have a faster innovation and training and time to market. So just, just very briefly, I would like to talk about what agents are. So these agents are on a spectrum. You've got the simple sort of chat with agents. These are the the basic retrieval agents. They are already generally available. You've got, they can retrieve information from ground and data. They can reason, summarize, answer user questions. You've got basic agents who can undertake tasks and, and do basic things, and you've got the fully autonomous agents who are doing things on their own. So, this agent is essentially a spectrum, but every agent is built up this exact same way. You've got a, you've got the orchestrator who's essentially the central nervous system of your agent, that can tap into knowledge, skills, autonomy, And, of course, these agents are built on, foundational large language models, such as the models from OpenAI or Anthropic. And what's, what's really important here is that not every agent will have a user experience. So these agents, user experience is optional. So there are agents that can run fully autonomously in the background, and the orchestrator of these agents can tap into knowledge sources, skills, and has autonomy so they can plan, manage exceptions, and do self learning. And what's important and what we're seeing emerging is these agents can talk to each other. They can trigger each other, so we are increasingly talking about an agentic ecosystem. So this is the intro. This is the foundational piece, and what's important for all of that is right. But how does this apply to field service, and how do we actually get started here? So field service is a is a really interesting and and and peculiar piece because you've got troops on the ground, people who are attending in person. So there's an adamant that AI will not be taken away in the foreseeable future as long as people will need to get behind the wheels and show up, to open a door or or attacking a screw. Right? So AI will not do that. So what are we talking about? What are the agentic opportunities for field service? First of all, what we're seeing is still a lot of companies are doing very many things on, manually on paper, technicians writing things down with with pen and paper, and then handing it over to an assistant at the end of the day, who would be keying in the data to invoice, to issue an invoice to the client or, write up a service or a work order. And needless to say, all that manual and and off system labor results and data inconsistency is prone to errors. So we want to get to a stage where we can, first off, introduce a system based on which we can start building our AI processes. So I'd like to think in terms of maturity models, and this applies for both field service and customer service. So a lot of the companies that we're stealing are still that are that we're seeing are still in the manual piece, manual labor. We want to make sure that we move them over to the managed piece where there's already a system, beneath them, and that system will yield proper data and and access to, to the right information. And once you've got that foundation laid out, you can start automating processes, and only after that can you start moving into an AI operated land. So let's let's see what AI readiness looks like at each stage and how AI can help every company move into the next, next step and what needs to be done to graduate into the next letter piece. So on on the manual, stage, in order for you to graduate, you want to implement, a work management system, so that you can digitize your paper based work and have a proper understanding of what you do. If you haven't got this, you might not even be able to chat with Copilot, but that that would be your AI baseline. And the AI opportunity here is you will have automated ticketing creation and automated summarization. And and that's what I showed last week in the customer service episode when a email comes in, AI picks it up, summarizes it, creates a, summary post into Teams channel. The human in the loop would pick it up and, click next, and that would actually create a case in your Dynamics or Power Platform instance. So that, if you do that, that will get you to the manage piece where you can actually start matching or mapping and tracking KPIs, SLAs. You can start building out a knowledge base. And you can at this stage, if you are not heavily investing into AI, you will default into searching and categorizing the AI use cases. But this is right for introducing human in the loop AI flows, and that will help you get into the automated stage where you can actually start putting and doing automated autonomous agents. And this is the stage where we want to get AI generated responses and getting a voice agent. And that's one of the things I'm I'm actively building out right now with my team is to have a voice agent where the technician just talks to the mobile app, tells the recorder number, what they've been working on. And based on the information that was dictated to the agent, the information gets parsed, and the recorder is updated in the system of record. So I was very keen on showing this, but it's not quite there to get it into the mobile state, but I will show you the backbone of that, today. So once you've got this, once you've got, you can have once you've got autonomous agents running, you might want to be considering moving into an AI operated land where you've got work orders scheduled automatically based on available capacity. You would have end to end resolutions. You can actually have, IoT enabled devices, and AI could start, start doing troubleshooting by itself, before scheduling human workforce, for on-site visits. So we are all doing all of this, obviously, to to increase customer experience and save manual labor and increase some of the KPIs that are most important for field services, which is first time fixed rate. And well, that's that that's the that's the golden one, first time fixed rate. But also making sure that we can adhere to SLAs. The technicians have the right skills and parts and everything, available to them to do the job necessary. So what I'd like to actually show you today is how we can move from manual to managed. And this is based on a true story, a very recent story. One of our SMB clients in Hungary, they have done a a license audit, and they realized that they're so small that at their level with three or four technicians, just getting paying for the field service licenses is not viable. They love Power Platform. They they have actually been using it. They have been getting value out of it. But just getting the you know, checking out the licenses, they're not getting the the impact that they are they are anticipating. So they are now considering moving onto onto an Excel based world. And what I'd like to show you is what I'm going to show them as well, how easy it is to build out a basic work order management system without all the fluff, without all the additional enterprise grade features that some of the largest companies in the world are actually using, but might not be required for small and medium businesses or even mid level companies. So I'll actually show you the back end, what's going on under the hood, and how you can build, ready to use power platform based applications with just very simple prompts and how you can build Copilot Studio agents on top of these and how you can actually check with your data. So what I'll show you is what Microsoft calls plans and different agents, the architect agent, the process agent. And let me pull that into the screen right now. There we go. So here, I'm zoomed in properly. I've got a very big screen in front of me. I'm in sort of the back end of the Power apps and I can just start creating a new plan. And I've got two environments, one where I have already set this up and if anything goes away as it could with live demos, I will just jump into the pre configured environment. But I'd like to show you a real live demo that that we are building as we are as we're speaking. So you'll see how important it is to understand your problems, your business need, to be able to articulate it. And once you do that, the solution will come quite naturally. So I will now articulate my business problem, and I will get to a stage where the agents would actually build me the data schema, the automations, and the app themselves. And I will be using the autonomous agent to create, demo data for me, and then I will show you how you can chat with your data once that's done. So once I describe my business problem, let me just show you what we will be getting to. We'll have this real clever flow, and we will have this work order management suite ready for us. So now I'll actually start building. So I'd like to build a work order management system for a heavy machinery services business where tech technicians service customer assets and track work on work orders and work order tasks. Right? So we are not actually using Dynamics three six five field service here. We are building everything from the ground up, and what you can see is it's describing the business problem, the purpose, this plan serves as a blueprint for addressing the business need, and it's giving me the user requirements. It's actually formulating user stories. So the technicians responsible for servicing as a user, I need to view sign work records, update, attach photos. I've got a service manager, I've got a customer representative, and let's crack on. That looks pretty good, right? I could have edited it, and it will give me now user processes, and it's building as we go. So you can see I've got one people present. I could bring in my colleagues, and we could be interacting on the same agent, on the same plan as it was a chemist. I've got the requirements agents, the process agent, the data agent, and the solution agent. So pretty clever stuff. I've got the user processes, written off, the work order initiation and assignment, and I've got the work order execution tracking. Let's check out one of these processes, process covers, yeah. Service technician, service manager. Actually, let's let's focus on the work order execution and tracking. So what we have here is a top down process flow, not too shabby for something that AI just wrote up in a matter of seconds, something that a BA would have spent a couple of hours doing. So we can actually see what's what's going on here, the different decision points, system sends completion notification to the customer, Customer manages monitor status, updates, reprioritize it, order. Right? That looks good. So the data agent is actually generating the data in the background. So I am building out assets, attachments, and you can see what I've already built has turned green, whereas the other things haven't. So I want to now check out my yeah. So I can actually edit the table and look at my data schema here. So notice that the ones that are green have already been created. And I could chat I could actually chat with this. These are the actual fields that I have. I could even add a new field if I if I wanted to. There you go. These are the different columns, and I would just keep it fairly simple and vanilla and go back and tell my data agent that this looks good, and my colleagues were asking me when I showed this to them, is this actually releasable? This is actually solution aware, and we've got the technician, task creator app, this would recommend to create a canvas app, but I would like to create a model driven app. The model driven. That's model driven. It could even build me a portal, a customer inquiry assistant, and a task notes helper agent. Oh, wow. That's clever. So let's build that out. Tables have been saved. I want to create all of these. So if I go back, I can actually now see where my how my solution is looking. So we've got the technician's task tracker. I definitely want that app to be created, and it's taking me to an app creator agent, and it's creating the app live as we speak. There we go. Let me just come back here and trigger off the manage recorder console as well. So that's another app that we will be building as it's getting things ready. Let's see let's see where we are with the with the app. Okay. So we've got the pages, recorder tasks, and we've got the user role service technician who can do this and that. So once we the technicians can view and manage technicians, manage for a quarter of tasks, what are the actual requirements. I noticed that here's a preview of my applications, and it's actually created demo data for me as well with, active technicians, work order tasks, and it manages attachments as well. And here I can define the different views, that I want my work order to have. Okay. So with this, I can actually publish my app. Notice that if there's anything that's not to my liking or if I wanted to add new new things, new fields, new views, I can I can do it? I can add a new column. I can add a new table. And if I'm building on top of an existing environment, then I can actually start building a brand new, reuse existing, tables and forms and and things that I have been doing. So even if this doesn't actually meet your needs hundred percent, this is a great accelerator, and you can expedite, the creation process significantly. Notice that I've started building this whole thing about ten minutes ago, and I already have my data model created, the requirements, the processes laid out, and two different applications have already been created. So let me just refresh here because it's disappeared. Alright. So I have now created the model driven apps. I sure hope they will come back. And and I've created the models. Yeah. So that's not showing as created yet. I think it's still in the process of deploying. And, essentially, that's that's where I wanted to get to. And let's just look at this overview of notice just from a single requirement actually built out service technician, service manager, customer service representative, personas, the different applications, the model driven app, site, PowerUp to make flow. I have an agent designed, two agents actually, and I have the whole data schema data model created down to the field level, and those were created in the dataverse as well, with all the standard fields. So it's nowadays have become this easy to actually build things. And let me show you the working example that I brought up, as part of the prep. So that would be the work order management suite. So here it's called users rather than the technicians. Here are the work orders that I have, and here are the customer assets. So I've got a work order Audi zero zero one, scheduled update. I've got a routine oil change under it, and, and I can track all the customer assets as I need to pass on when it loads. Cool. So as I said, this is all well and good. I'm actually using AI to get from the manual to the managed level. But now I would like to get to a stage where things are automated. So if I jump over to Copilot studio and I've built a field service agent, it's a recorder. Let me just close that. Notice that I have a very, very basic agent. All I have added is just a Microsoft Dataverse connector. No special sub agents, no knowledge, not even any instructions. So this is as vanilla as it can get. And if I start a brand new session let's look at how clever this is. Let's give me the assets for Audi. Excuse me. Oh, jeez. Right. So give me the assets for Audi. And here it's actually sending off and triggering the Dataverse MCP server, which is in preview and model context protocol is essentially a way for AI to to interact with with data. You can see that it's writing the queries as we go. And, obviously, your users and your customers wouldn't be interacting with this directly. This can be deployed into Teams or a mobile app or SharePoint or even in a voice mode. So we are very much in a in a test phase now, and it has pulled in all the assets that belong under Audi, the a four set on. There are five assets registered under the customer Audi, all classified as vehicle type. Is there anything else? Let's actually create a new asset. Please also add an a eight l sedan. And it will now start creating an asset, and I could give it the serial numbers and so forth. And it's actually doing the record creation. So notice that as we discussed, the agent is calling on the tools that are made available. So it had a describe table, read query, now it's pulling a a create record. And I have just created that new record. So imagine if I work at a, car shop and someone called Rosie with their new auto Audi and they want that serviced. I need to register it. I can just chat with my system, and the record gets created. And I will now ask it to, please create a work order for a annual inspection for this a eight o. That's it. So when we talked about the fact that Copilot is the UI for AI and it is becoming the overarching wrapper of classic or legacy or, traditional systems, this is what we meant. We are I no longer need to click anything. I can just come in and type things out. And let's see what was created. Right? Work order twenty twenty six point zero zero one. And let's check out that new Audi a a sedan that was just created, and I can check out the related work orders beneath it zero zero one. And that one has the annual impact inspection comprehensive vehicle check. Okay. So let's say someone has completed that work order. Please set work order to completed. Okay. So recorder has been updated. Let's check that out. I would expect the status to change from open to completed, and it did exactly that. So this is the new way of building and interacting with applications. And imagine this can be deployed to a mobile user, and they can even chat with it through WhatsApp or Direct Line speech, so they can just talk to the system and and even third party applications like Salesforce service now, which are both heavily used by field personnel, can be used. And, and this is this is the way forward for, for technicians and field service folks. So I hope you found today's demo helpful, and and we really showed, in a matter of less than half an hour twenty minutes how you can move from manual to managed and how you can get into an automated world, we have now created a practically user interface on top of a, a business system, system of records that people can just very very straightforward interact with. And let me talk a bit about what we're seeing, why people are not doing this. Right? If this takes a matter of twenty minutes, why why isn't everyone doing this? So people say that AI is easy. As you could see, I just did this in twenty minutes. And it is easy if you have the know how, if you have the processes, if you have the right data, if you have the understanding of how to do this. So that means if the organization is ready, AI is actually easy, but they need to get there. And another common one is we need to sort out everything first. But in reality, you just need to get the basics first, and then you can get going. The same way I just build this app from scratch, you can start building something and then start extending and building on top of it and and moving forward. And the fact that you have Copilot because you have an AI assistant doesn't mean you have an AI agent. Therefore, that won't be a true transformation. So the actual agents that I showed you that enabled me, to chat with my data and update things, that's where true transformation happens, and that's where we want to get to with a proper system of record. Now AI is expensive and it hallucinates. Notice that I just created all these records without an issue. I updated them. It retrieved the right data. If you paid attention to the serial number and all that, there was no hallucination at all. Hallucination happens if you start to misuse AI. If you just give it too much context, if if you expect too much of it, then it will start hallucinating. The other thing is data isn't ready. It's a common objection. It is a real blocker. But that's where we want to hone in on the right use cases and making sure that you are identifying that thin line of of stable layer that you can you can actually just get going with. And if you sort out those critical bits, it will be pretty straightforward to get the basics right. And the other piece is people won't adopt it. And I say two things here. People will adopt AI, if it solves their pain. Imagine the the chat app that I showed you. It can save them from so much handwriting, messy messy papers getting lost. It's sort a real pain, or if you come up with a autonomous agent that you might not actually need to adopt it. So that makes it all the easier. And what we're seeing is that companies who start building agents now, you'll combine their knowledge and, similar to how you put money in the bank early on, it will compound, like, interest, and and you can start getting value from it earlier and just always start building on top of what you have. And this is exactly why we came up with our value led methodology where we work with our clients to identify the value first, lay that on top of the their understand how they create value, what's their current architecture, and what agents land themselves, to be built, and then we can move over to execution. And what we built out is a one day in person, workshop with our clients where we have every decision maker in the room, and really hone in on value architecture and design, And we will be building an ready to execute road map so that you in ninety days, you can get to a stage where you can confidently say that we are getting actual business value from AI. And what we're seeing is companies and people who win in the AI race aren't the ones who are the smartest or the best. The companies who are winning in the AI race are the ones who are starting, who are experimenting, but the ones who are actually getting proper business value out of AI. And the earlier you start, the earlier you the earlier you will reap those benefits. So we can help you accelerate this journey by identifying the right use cases, honing in on the right foundational pieces, and then if you like, we can help you with the execution. If you scan this QR code, you can go to our workshop dot vision labs dot com website, where if you register, I will personally call you, so that we can figure out if we can help you. Thank you so much for attending today's session. Hope you found it useful and helpful. And next week, we will be talking about procurement. So from the CRM related workshops, we are moving over to the ERP sessions, and procurement will be all about analyzing vendors and, and understanding our spend data. So thank you so much. Hope to see you next week. Bye bye.