AI Readiness and Value-First Thinking

Alright. Alright. Hi, everyone. Welcome to this year's first webinar. This will be about AI readiness in ninety days. Okay. So here's the new platform, and we'll just wait a minute, and we'll get it going in a second. So okeydoke. We've got people flowing in, so we will wait for a couple more minutes. Okay. And here in Hungary, we say we've got the academic three minutes delay. So we'll get started. First of all, happy New Year, everyone. And let's get going. So this won't exactly be a recap of the of the webinar that we held in December, but it will be very similar to that one. So if you've been been there, quite a few things might sound similar. What I will extend on is with our team, we held a two day AI hackathon with Microsoft Copilot Studio. We have quite a lot of learnings there. So I will want to give you some insight into into that, and and hope you'll you'll find that valuable towards the end of the session. So let's let's talk about AI readiness and and AI readiness in ninety days. So we firmly believe that in twenty twenty six twenty twenty six is the last year when it's not too late to jump on the AI bandwagon, and you can actually start reaping the benefits of AI in the first quarter, but not in a way that most companies would expect. We we see a repeating pattern is that companies need to get ready before they can really start using benefits. So let me start talking about a couple of couple of these things. So a lot of people think AI is just really easy. We can just, you know, use Jet GPT and Copilot and Claude and Gemini or whatever and just go go crazy with it. However, AI is only really easy, if your organization is ready for it. Otherwise, it's a real, it's a real pain to use or you don't really get the benefit that you really look to get out of it. Other the other thing is quite a few companies believe that AI must be perfect. And that's not really the case. You just need to get going and start trying and start learning as you as you go. And that learning will compound as as you go ahead. And I will talk about that in a in a minute. So another one that we see is, we are already well underway on our AI journey because we use Copilot. Well, That's actually one of the things that Copilot is not AI transformation. Copilot is just chatting with someone that will give you personal efficiency, but in reality, agents are the ones that will give you proper business value and and will get you that AI transformation rather than Copilot. So Microsoft have done this study where they looked at companies who have done their their homework and did their readiness piece, And those companies were building agents two point five times faster than the companies who who didn't do the readiness piece. And another one is that these companies start compounding, getting the benefits way easier, and they just stack up similar to how you put a lot of money in the bank, and it will start paying interest on the interest. And that is the beauty of compounding. So if you start earlier, you will have a lot bigger advantage compared to the ones who are starting late. And this is why I say that twenty twenty six is the last year when it's not too late to get started because compounding happens so fast that in a year or two's time, the companies who start this year will be way ahead of the of the laggards. Now this is what we'll be talking about today. We'll clarify some key terminologies. We'll look at the main failures of AI initiatives. We look at our model that Visual Labs have come up with the V shape model, and we look at a couple of very tactical practical bits, how you can get started and what could be the next step for you. So hi, everyone. I am Balaj Horvath. I am the founder of Visual Labs, and I've been working with Microsoft technologies all my professional life. I've started with Microsoft Dynamics ERPs. And from there, I expanded into Dynamics CRM power platform and data. And I'm proud to have built Visual Labs, which is a company that works only with Microsoft business applications. This is our wall of pride. These are the certificates of all our all our colleagues. And on the right, you can see all of our badges that we got from Microsoft as a proof that we are good at what we do. And these are our clients all across Europe and the world, and we've been quite heavily been expanding outside of Europe. We are primarily based based in Hungary. That's where our head office is. About fifty percent of our revenue actually comes from the UK and outside of outside of the country. So today, I'd like to give you an introduction into the VAD methodology, which is value architecture design. And the one of the things that it sets it apart from from many methodologies is that it starts with value. Yeah. We very much encourage our users, clients to to think in terms of business value. How can you get value out of a use case of an investment of an initiative? And only after that, only after you've identified that it makes sense doing should you start thinking about architecture and design. So that is the cadence. That's why it's called VAD, value architecture and design. And as part of the next piece, let's clarify some of the some of the the key terminologies that that we'll be using, and you might hear flying around. And I should mention that this is the first episode of a webinar series. That's why it's important that we lay the groundwork and the foundation so that we can start using these these terminologies confidently without having to repeat every single time. So the first one that Microsoft have really started using in the past past year or so is Frontier Firm, which is an organization that is human led, but AI operated, where AI agents work alongside human workers, people to run the core business processes. So in a couple years time, most companies will be frontier firms, but now they're like pioneers. That's why they are the frontier of of how an organization works. And Microsoft have this frontier program, and this is what we do as as our mission. We help companies to become frontier firms. But in reality, it means that you are ready for AI. You can you can restart reaping the benefits of AI. Now the other big thing that has been flying around is Copilot. Copilot has dual meaning. Outside of the Microsoft ecosystem, Copilot is a worker or an AI assistant that helps you be more efficient. Anything that that is working alongside a human is a Copilot. And in the Microsoft world, it's it's the brand for AI. Right? So everything is a is a is Copilot that has AI in it. I do think that Microsoft has been misusing this terminology, to be honest, but but it's it's part of their positioning that Microsoft says Copilot is the user interface, the UI for AI. That's why it's all over the place. So you'll you'll see us using Copilot quite a lot. And then let's talk about agents. Agents is was the word of of last year. They have they have really become a mainstream phenomenon. And the word agent has been very much misused across the the the the year last year. And as a matter of fact, in the next week's episode, we'll be about the agentic technologies within the Microsoft stack. I will be work walking you through six different pieces of technologies within the Microsoft stack stack that can be used to build agents even for a business person without having to write any code just through prompting. You will have a working agent, but it's it's all over the place now. So I really want to make sure that there's some clarity around how how what can be used to build agents. But before we jump into that, let's again clarify what agents are. So agents are AI assistants designed to automate and execute business processes working with or for a human. So this is very bright on on the big screen here. But essentially, an agent has an orchestrator that can tap into knowledges, scales, autonomy, and is built on a foundational model such as OpenAI's. The latest one is ChatGPT, or sorry, GPT five point two, or it can be on Claude, or it can be on other other foundational models. And knowledge may be something I can upload a PDF or a text doc, or it can be something like a whole book or a whole database. That's what it's been grounded on. It can have triggers and workflows, so it can initiate or it can be initiated, let's say, email comes in or something happens in a in a system or system record or some someone hits a button and an agent is triggered, and it can actually, plan its own actions. It can reason. It can handle exceptions. It can do self learning. And this is really important that not every agent might have a user experience part. So many people think of agents just as a chatbot, but in reality, there are autonomous agents that haven't got a user interface, user experience. They just work in the background in the back end without interacting with anyone. And these agents can be connected, so we are increasingly talking about multi agentic world. And that's that's that's a key part how these agents are interacting with each other, how they can call on each other, what functions they do. And we are increasingly seeing, like, a master agent or an orchestrator agent who would be who would be calling, receiving, and handling information across these agents. So it really becomes a network of agents. And that's why it's so important to to get these right. And and that's how you can really get transfer AI transformative benefits. Right. So so much about the definition of agents. And let's talk about a bit about the categorization of the different types of agents. So you've got the very straightforward basic agents, which are sort of chat with your data, chat with your file. Those are the retrieval agents. They can retrieve information, from a from a certain doc. So this this is the one that most of us are using. As a matter of fact, LaxiFetch, GPT, and Copilot, they are all agentic large language models because they can, you know, reach out to the Internet. That is actually a tool. They can retrieve information. And there are agents that can perform tasks. So so even if you are you just using ChatGPT or Copilot and you have a connector and you wanted to do you know, send me an email and it's actually connected to your Outlook or Gmail, it will it will write an email and send an email for you. So that is a task based agent. And then there are the autonomous agents, which is where we want to get to where agents just get triggered, they handle something, and that they just started move move forward and and and actually work work alongside a a human worker even if an when a when a human is is off sleeping, an agent can do his thing. That's why it's this is this is the key. This is what we're shooting for. Now let's talk about why it's so important to start building these agents and and and where the benefit and the opportunities in the agentic world. So in a legacy traditional organization, so not a frontier firm, but the companies that we are currently working in, our current environment, we've got capacity gaps, knowledge gaps. You know, we've got certain amount of people with certain amount of knowledge. Not everybody knows everything. You know, we all know that one person who has access to that knowledge, who knows how to do the month end process. And if he or she is on leave, then we are in trouble. So they become bottlenecks. And we are grounded by how fast we can work. And there's slow, tedious manual processes. And let's be honest, every job has areas that is boring and repetitive, and humans don't like boring and repetitive stuff. So they are best to be automated. And if something becomes boring and repetitive, the quality tends to be shifting. We quite often won't have an access to do to do things. So in today's world, all across the globe, but especially this this this part of Europe, well, even in the UK, you've got increasing rising costs, inflationary pressure is mounting. So everybody's very much pressed, pressed on on cost areas. And all of this results in slowing agility. And now let's look at a frontier firm. So if you've got agents in your organization or working alongside your human workforce, then you have scalable digital workforce. Imagine you've got an influx of of a number of incoming calls, or issues are flowing in, or you just launched a product and and you've got all these sales orders that need to be handled, you can just multiply the number of agents instances that are running, and you can just scale up. Whereas you can just hire more and more people of the Infinite. And and even if it's just a daily spike, then then you are just Preston workers. And with the agent tech world, you can really build out this enterprise knowledge engine and remove the human bottlenecks, you know, that one person that you rely on to do the month end. If you can just obtain that knowledge from them, then it's it's a big win. So you've got automated workflows and processes that operate outside of your regular nine to five. You've got human agent collaboration, so it can be a back and forth between between workers and and agents. And you can really just get rid of the low level tedious task and really rely on the human side of things that require creativity, humor, customer experience, you know, that should get all the attention, not the back end tedious stuff. And agents will provide consistent high quality execution, provided you build them right. And all of this will enable faster innovation and faster time to market for industries and clients where it's where it's paramount. So this this is one diagram that I've saw I've seen, was it almost two years ago, now a year and a half ago, about Microsoft's vision of the whole agentic organization, and you can see on the left hand side Copilot chat, and that Copilot icon really being the UI for all the back end AI stuff. And you've got all the AI elements, Copilot, agents. And what's increasingly important is that you've got this whole tech stack where you can build your own agents, whether with a partner or whether you build the in house skills. And you can actually build agents on top of your legacy existing systems, even if they are not Microsoft products. So if you have Salesforce, ServiceNow, SAP, or Oracle, you can use the Microsoft stack to build agents on top of it. And your workers, your users will have an option to interact with these systems through agents, through Copilot, or directly go into those systems. And that way, AI really becomes a whole new way of interacting with your systems and interacting with with the agentic world. So it's all well and good. It's a it's a bit of an abstract diagram. So this is why I created our own way of thinking of organizations. Again, this left to right notion. But if you think about an organization's digital landscape, every organization needs to have a communication layer, how people interact with each other. Increasingly since COVID, this is more and more in the digital world, So we are using using Outlook and Teams to communicate via even if it's in via calls or chat or emails. You've got a document layer. We are storing your our documentation on on SharePoint and OneDrive and Loop and and whatnot. So that's where we build out that knowledge. And we've got the application layer, which may be Microsoft based or or not. So you might have an SAP instance, Salesforce, or you might be using Dynamics, you might build your own apps with PowerApp, and that becomes your app layer. And all of these constitute an organization's data layer. And what really happens with AI is AI sits on top of all of this. So AI is just another layer on what you already have. And if you have a tidy room, then AI will be will be nice and clean. It will be able to work and understand your your landscape. So we've got, as I already mentioned, this is the the main part of next week's, episode as the tools to build agents. So you've got this AI layer, and you can use m three sixty five Copilot, SharePoint, Fabric, Copilot Studio, Foundry, and you can even build a coding a a co yeah. Sorry. Agents with a software development kit, but we won't go into too much detail. The purpose here is for you to understand that you can link up this whole ecosystem just by staying in the Microsoft ecosystem and using all the tools that are available to you quite often, not even requiring additional license costs. So this is the landscape. This is the complexity that we live in. This is our reality in twenty twenty five. And now going into twenty twenty six, we can really start building that AI layer because the technology is is is there. It's it's ready for it to to be used. So let's talk about what it means to be ready, what it means to be AI ready within the Microsoft framework. And in short, getting that data layer right, getting that communication document layer, app layer sorted is what will give you the the most benefit of your AI layer. So just, you know, that's the one takeaway for today. There is a Microsoft framework for agentic readiness. You can scan the QR code and really get that seventy odd page PDF, but I will spare you the time of reading all of that. So here you can see the business senate, and here you can see the main part of that outcome. Bear with if I may have audio issues. Let me see if that's coming back. Connection now seems to be okay. So right. So let me just recap. Okay. So let me just recap a little bit. So this is Microsoft's AI Readiness Framework that consists of of the five pillars, two major sections. You you can talk about the strategic readiness and the execution readiness. Right? So strategic readiness is whether you have a overarching business and AI strategy. You are not just working in silos. You know how you want to build agents, what's accepted, what's not accepted. And you've got the business process mapping. So if you if your people are just working in an ad hoc manner, don't expect AI to come in and do the work for them and instead of them, because you need to get the process mapping sorted. So that's the the strategic piece, and then there's the execution part. You need to have the technology and data in place. As I already mentioned, the data is is is of paramount importance, but you also need to have the skill set and the people side of your organization, and your people need to be ready for it from a cultural perspective. They want to be innovating. They want to be trying out all these new technologies. And of course, as with any other IT product, we need to be talking about security governance, how we're going to be building this out in a scalable, safe way. And don't get me wrong. With AI, you are opening up a whole new area of areas of vulnerability, which can be a a total totally different subject area. So this is why Microsoft is a is a very strong offering, because a lot of that security and governance comes baked in to the product. So these are the five main areas. And then every one of those five areas have subsections, which I won't go into too much detail now, but you can just you can just really really see it here. Cool. So let me just probably zoom in a in a in a little bit so you can can see that. Alright. So moving on to to the next piece. Common pitfalls and success factors. What we see, why quite often these these AI initiatives don't work. And according to the Microsoft study, if there's no AI vision, if the teams are working in a siloed manner, if you haven't got value defined, and this is why it's what's really important for for our perspective, is getting value identified and just make sure that you know what you're trying to get out of it. Otherwise, it's just gonna be learning for the sake of learning, which wanna get you proper business results. What we're expecting in twenty twenty six compared to twenty five is executives and boards will be expecting tangible outcomes of all these AI initiatives. So now it's no longer, oh, let's just do say AI for the sake of AI. Let's do AI for the sake of business value. Immature data, I can't stress this enough that you will not get to where you want to go if you cannot feed AI with your contacts, with your data that's specific to you. And usually, the main reason of an immature data layer is a weak architecture. If your systems are not connected, not maintained, or you don't understand what systems are even being used if you're using five different CRMs across the different departments, then you you will not get to a right data data part. And also, you want AI to just do too much, if you're expecting it to substitute your whole workforce, you'll never get there. So make sure you set realistic expectations. And let me just show you, as part of the next piece, what it is that we're doing to to help our clients. And this is something that helps you think in a in a value driven manner as well. So this is our our our methodology proprietary methodology for the value architecture and design. And this is how how I recommend approaching all AI initiatives, is let's understand how you create value today. You can do this through process mapping or creating a value proposition canvas or value stream mapping. There are all these tools out there that help you identify how you create value. And from from that, you can then move on if you understand the main levers of your organization. What is the underpinning IT architecture and organizational architecture that supports that value creation? And once you understand the lay of the land of your legacy world, you can really start designing the transformation. How is it that you'll get to the next stage? And what architecture changes are required to really move on? And what will be the new pattern of value creation? How you'll get there? So this is that V model value at the beginning value at the end, and then design bang in the middle. That's when the messy, messy deep, and that's where the transformation happens, and you'll emerge as a as a frontier organization. And this model works because it really gives you clarity momentum, and you'll get to compounding pretty quickly in a way that you can right size the experiences, experiments, and you can really get to making sure that you can start delivering value in a quarter. So the reason why we're setting out this program, our program in a way that you have one day of planning, because if you cannot plan it in one day, then there's no way that you can execute it in ninety days. Right? So so that's a that's a pretty good proportion of let's do a focus planning day and then for for the next ninety days of execution. That's why we are we like to think in a in a ninety day cadence. Cool. So there are quite a few objections that we hear as we as we go and as as we talk about AI readiness. Many people say that AI is just really easy. It's just everybody's can use now chat GPT, they just go into chat dot open AI and then or Gemini or Copilot or whatever, and everybody's an AI user, which is right. But you will not get true AI, transformational value out of it unless your organization is is ready for it. So once you lay the foundations right, once you become AI ready, that's when AI becomes really easy as well, because all the use cases will sort of magic magically be unlocked. Right? Overnight success once in the making. And we need to we need to get the get everything right first. And it's quite easy to fall into the construction industry trap with IT, that people think that everything needs to be planned out up to the last brick and turnkey and everything, and only then can we start building. Whereas in reality, one can actually start to build and and get things right once you have the right foundations in place. I quite like to use the analogy, especially in this cold winter, as you don't need the whole lake to be frozen over. You just need a very thin path where you can safely cross and you will get to the other side if you know your way around where the ice won't break. Now another common one, which is already mentioned, is people look at Copilot as a way of transforming an organization and bringing into the new frontier world. But in reality, AI with Copilot won't be a true transformation. Agents will get you the true transformation. And still very common one, and it is true, AI is expensive and hallucinates. And that's why it's so important to find the right tool for the right job. So don't misuse it. Make sure that you're giving it a small enough job or you break up the pieces of job into large enough chunks, so that it cannot hallucinate and it's reasonably sized and it's reasonably costed. In twenty twenty six, I expect that there will be an emerging token optimization, like token engineering alongside the context engineering phenomenon that we're we're starting to see. So how you can right size the prompts, how you can make sure that there's no hallucination, I think that will be that will be a key topic for twenty twenty six. Yeah. Our data is isn't ready. Most likely is it isn't. And you will never get it perfectly right. But make sure that you get the critical bit sorted. For instance, you need to know your customer base, how many customers have you got, your products, your inventory, you know, if you want to, if you want to go into a certain area, just make sure that you got that sorted. Our people want to adopt it. And there's two things there. They might not even need to adopt it. If you are building truly autonomous agents that just run-in the background without human interaction, that's beautiful, because people don't need to adopt anything. AI just works in the background. And the other one is if you are truly solving proper pain points, then people will actually like it, right? They if you're solving a problem that they currently face, they will be welcoming that. So just make sure that you're finding the right problem to solve, and you'll be greeted with open arms. So those were the common common objections and how we usually respond to those. I definitely recommend that as you go through and as you start your AI initiatives, or as you want to bring, take them to the next level, just keep using a lot of these and just be prepared if anyone comes up. And another one is, how do you get actually started? So as I already mentioned, having a one day planning workshop, where you really hone in on the value, the architecture, and you come up with an agent backlog, then you just start executing. Just make sure that it's timed in and and it's set up right. And you have all the decision makers in the room. That's what we see is everybody wants to get started with AI and they they start creating all these initiatives, these big AI roadmaps and strategies and whatnot. But the technology has been moving so fast, even in the past three months, that whatever stood in October, November time, there's now a whole different approach how you can you can do it approach it. If you had done it then in a couple months time, you've already been reaping the benefits. And now you would have something to build upon to further improve. But if you came up with a plan and you didn't execute it, now that plan is most likely obsolete. So I would definitely advise you to split it into smaller chunks and and start thinking smaller pieces. Because this is and this is key. The companies who start early will start compounding. I know I had this slide, but I think this this is the one that really brings it home. Why twenty twenty six is the last year when it's not too late to get started with with AI, with agentic transformation. And what I'd like to do in in the first quarter, right, in the in the in the next ninety days is help you find the right areas and the right use cases on how you can really make benefit of of AI specific to your department. So next week, we'll talk about it will be more of an introductory session about the different technologies. I will actually be giving you live demos of every one of these these pieces of text. Don't worry. There won't be technical deep dives. It will be just flashing up, showing what it can do, showing the license implications, showing the best use cases, so you have an understanding of what what it is that we are facing. And then the week after next week, we will actually be going into departmental sessions. So in the first four sessions, we'll be covering in the first four sessions, we will be covering marketing, sales, customer service, and field service. So more customer related items, more CRM stuff. We will be looking at the out of box agents that Microsoft provides and what are the most common use cases that can be that can be implemented and that have been proven already as good solid use cases for AI. And then after the CRM areas, after field service, we'll cover supply chain, finance, HR, and IT. And those are more sorry, ERP back office pieces. And only after that, we will be going to sort of the keystone, the the capstone piece that will cover cover it all, close it all, which is agentic and AI governance within organization, how you can make sure that those AI agents are operating in a safe and secure way with the necessary oversight and governance that's needed. So with if you follow along with these sessions, you will have a pretty good understanding of what AI can deliver in every one of these departmental areas. But also, I totally understand if that might be, you know, too overwhelming and too much too much of a commitment to join in. So I would recommend to to come into the sessions that will give you the that is most relevant to your department. Right? So if you are in the sales function, sales marketing, customer service might be relevant to you. If you are in finance, there will be a session about finance, supply chain and a r HR might might be relevant to you. If you're in IT, there will be an IT and then the governance session as well. So what we'll also do is on the back of these sessions, we will be creating agentic white papers and and and fact sheets on how you can get started. So please do do come to those. And if you wanted to work with us, if you wanted us to help you find the right use cases and execute either execute for you or help you execute and then teach you along the way, then we are offering this one day AI readiness workshop where we go through that exact methodology of understanding how you create value, what is your current architecture, and then we'll be designing that feature. And you can scan that QR code that will take you to the to the registration form. And if you register, I will call you today or tomorrow to start discussing the next steps to make sure that we can we can help you out. So that was it for today. That was episode one of the twelve part AI readiness series. Next week, we'll be talking about AI toolkits, and then we'll start get started with the departmental series with marketing. So thanks so much everyone for attending the session. And as similar to the last time, there actually will not be a q and a part, a q and a session, because what we are finding is everybody has such different, such a such a unique position that if, you know, just in this q and a, I probably couldn't give you justice and understand your total landscape. And I don't want to go into into haphazard haphazard advices. But if you're curious, do feel free to sign up to the workshop, and and I'll give you a call, and then and then so we can start discussing the lay of the land and what you have top of mind. So alright. Thank you so much for for joining. Thank you for your time, and I hope to see you next week for the toolkit session and then for the marketing one the week after. Have a good one, everyone. Good afternoon. Bye bye.