The Enterprise Edge - Georg Glantschnig, CVP Dynamics 365 Agentic ERP, Microsoft

Speaker A: Foreign. Welcome, um, to another edition of the Enterprise Edge podcast. This is Mark Vigoroso, CEO and founder of the Enterprise Edge. Very pleased to be with you this Monday with the various team guests we have today from Microsoft. Georg Glashnig is with us from Dynamics 365 and he is the CVP looking after a Gentic ERP. Uh, Georg will have great things to share about not only where Microsoft is headed but also what he hears from customers across all different industries and markets regarding how they are making decisions regarding implementing some of the emerging technologies that are available to companies today, uh, across the sort of the agentic landscape which is moving very rapidly to say the least. But before we get into any of that, please join me in giving Georg a proper welcome. Georg, thanks for being with us today.

Speaker B: Thank you Mark. Happy to be here.

Speaker A: Great, great. Glad to have you. Well, let's get, let's get into it. You know, as we often do for folks who listen to this podcast on a regular basis, we'd like to get to know Georg a little bit. Uh, we'll start off with a, uh, couple icebreakers and then jump into some uh, market questions and product questions. Uh, and then we'll end with a bit of a, a speed round to ah, to end the session. So let's, let's get to it. You know, Gay Org has a, you know, you've got a great history, you've been around enterprise tech for a long time. Um, and uh, you know, your path has taken to where you are now, uh, leading a key piece of where Microsoft is headed regarding erp. Um, you know, you started back in the day at SAP, you know, with R3 and in the trenches with customers. Um, and obviously you know, a lot has changed since then. But I'm wondering in those early days, you know, we oftentimes look back early in our careers on lessons that we learned and perhaps tough lessons and I wonder, are there any of those tough lessons from your early consulting days that, that really serve you well even today in your current job?

Speaker B: Mark was a long, long, long time ago. As if I, if I remember correctly, I think 97, you know, I started as a consultant as you mentioned.

Speaker A: Yeah.

Speaker B: And Nokia was my first project. So I spent I think uh, a year in Helsinki headquarters creating the first global ERP template, you know, to roll out all the subsidiaries. And I, I think the most valuable thing I, I learned basically it was not about the software application, it was about the troubleshooting. Uh, as consultants you quickly learn that the first problem you see is really the Real problem, whatever you tell you something, it's not working, it's doing the wrong things. But then you look at system logs and tell you a completely different story. Uh, normally the root cause is somewhere in the middle and I, I still use exactly a similar approach today. Whether we are discussing AI strategies, any kind of modernization efforts, or if you're leading large organizations, I spend far more time in trying to understand the problem than just jumping to conclusions. Technology changes every few years and it's getting even faster if you look at the last couple of years. But I think good problem solving, dozens. So at least that's still something I'm trying to keep him.

Speaker A: Yeah, uh, that's good wisdom for sure. For sure. And you know, and, and you've ascended to a key role of leadership. And leadership is one of those things that I think oftentimes people assume you just sort of pick up along the way in terms of being good leader and um, in others I've observed and myself try to practice that. It's really a discipline unto itself and ah, skill set unto itself. And I'm curious, in your role as you lead teams there at Microsoft, you know, others have described you as someone who's very good at helping people and teams understand sort of what's behind decisions that are, that are being made, even if there isn't 100% agreement in those decisions. Right. And you never have 100% agreement. But there's still this, you know, how do you get everybody on board and rowing in the same direction regardless of that level of non unanimous agreement? And so I'm curious, how do you approach that with so many players at Microsoft? As fast as this market is moving, as fast as decisions are being made, as fast as products are being released, how do you marshal that unity, uh, uh, amongst your team members in this day and age?

Speaker B: It's a really, it's a really good point. Specifically as you said, you know, things are changing even faster and decisions are getting even more complex. And um, maybe the easiest answer is, you know, because I'm Austrian, um, I'm saying it's because you know, Austria is a small country, as you know. Ah. And I think at least one thing, at least I like about our culture is that we generally, we're not walking into a room assuming we are the smartest persons there. Um, and I think there's also influence my leadership maybe more than all these management trainings and courses you're doing. So my job isn't to convince everyone that I'm right. It's more like My job is to make sure that we have the uh, best possible discussion before making a decision and then basically make sure everybody understands why we made the decision. So I think I have learned over the years that understanding creates commitment. Agreement of course is nice, but it's, I don't think it's actually required. So if people at least understand the reason, know their voice was heard and see that the decision is consistent with, you know, our value and our strategies, I think that's usually rally people behind it and you know, you can go into pretty fast committed um, execution.

Speaker A: Right. Yeah, it's sort of art and science in uh, my experience and I think that makes a lot of sense and it is somewhat made more difficult given just the speed and the velocity that decisions are getting made. Right. And even with large organizations like Microsoft. So that's great insight. We can't all be Austrian, but we all aspire to be right.

Speaker B: At least in soca we would need more Austrians to be the stronger team.

Speaker A: Exactly, exactly. Well that's great. Well, let's garg. Let's get into some specifics about where you are at with uh, Microsoft Dynamics 365 and specifically around sort of strategy within the ERP scope when it comes to AI innovation, agenta capabilities and whatnot. And I think one of the topics that is um, keeps coming up in my travels and my interactions with basically everyone is as this sort of notion of an autonomous enterprise workflow process. Right. That, you know, where does liability stand? Right. When you have actors that are human and agentic and you have this sort of mix of AI operated, human led, uh, human in the loop. Right. This sort of hybrid of, you know, agentic and human workflow. And um, how do you manage from as a product leader and a business leader, things like governance, like where, where are the, where is, are there new lines that need to be drawn with regards to governance and compliance and liability. Uh, when you have um, workflows that are taking action autonomously. Right. With machines as opposed to humans and obviously combinations. So long way of asking that question around where does liability land if something goes wrong?

Speaker B: I mean it's a super important but also complex question. We have learned really a lot. We try to be very early in the market. Exactly. To just learn from our customers what's working, what's important. So let me break it down maybe a couple of different aspects and the first one at least, which is still fascinating for me is what we really learned is that people are far more willing to forgive a human error than an AI error. So When a colleague makes a mistake, which happens every day, many times, they normally get a second chance. But when AI makes one, the whole trust disappears instantly and everything is falling apart. And I feel a bit similar if you follow the autonomous driving conversations as at least in San Francisco, at least I felt it's pretty awesome how these cars already maneuvering and I mean human drivers are involved in thousands of accidents every day just in us. Uh, but yet a single crash or installment involving an autonomous vehicle often dominates the headlines as I think the expectations for AI are fundamentally different and we expect this kind of nearly flawless execution. And as you said now mapping this to the ERP market of course makes this even stronger. Uh, because the simple, there's no tolerance for these hallucinations as again we're talking about finance or we talk about purchase orders, they are black and white. I mean there is no gray in finance. Numbers don't lie, is the same. So the consequences, they are pretty significant. Uh, so in, in marketing, if AI gets it wrong, a customer gets a wrong email or an erp, you misstate your books. Uh, these are huge things. And that's exactly why agents in erp, we have to put them inside what we call these guardrails. Of course a single agent can look across ERP data. It can be planning outputs, it can be supplier emails, it can be, you know, policy documents, it can be conversations, you know, um, employees had in, in teams. Um, and they can basically draw these coherent pictures of, you know, what's basically happening and what to do about it. And I think some of these logical steps and actions that Asians can take, you know, they can basically take on this, I call it the routine work, you know, the runs inside your guardrails automatically. And they think about these guardrails basically in two packets. So uh, to your point, you really have to make this very, very predictable and secure, uh, specifically in the ERP context. So I think we have these two things. The first one, we have the ERP business logic. These are all these rules, these regulations and finance controls that basically live in your ERP system. So I'm a strong believer these ERP systems are not going anywhere because you need this grounding. There's no, at least I don't feel there's any reasonable explanation. Why would you rebuild these deterministic rules on a non deterministic layer, uh, at least in the next couple of years? I just don't see the reason. And what we are using basically is the MCP server. I assume maybe the listener have the context there that's basically how agents can communicate with your system. It's, uh, a communications protocol. So we use this MCP server. So basically agents work the systems and basically with the person. Then they can basically not break beyond this foundational logic because we are exposing it. And this is not what an agent basically is making up as it's working in these boundaries of the business logic. Then the second thing we have also the company policies, these rules your business is setting, they can live in this agent layer or it can be even a SharePoint document, as many policy documents in a company are. And basically agents can, uh, reference these documents. So if, for example, if an agent finds an alternative supplier, uh, because the supplier maybe delayed, uh, a commitment, if it's within a certain price threshold, uh, it can proceed basically on its own. And of course beyond that, the agent itself can bring a person in, as we call a human loop. But of course give the context. These are the facts, these are decisions to make. And then basically humans, uh, saying, uh, left or right. So I think that agents can even find in this unstructured information, as we call work iq, as we understand who is communicating with whom. What are documents in this unstructured world? You can also find maybe situations in the past where decisions were made. And again, everything can be outlined and presented to humans, uh, to basically proceed with these decisions. I think accountability never moves to the agent, uh, because every decision leaves a trace. So you always know what agents are doing, which of course actually gives you a much clearer way of understanding what's happening, who is doing it. ERP was always about audibility, about traces. So I feel more like we are trying to move people who are, uh, I feel more like the middleware. You're kind of between your system and the business process. Uh, we're moving people more into, I call it like the agent boss who is directing agents to do things. But again, they're just preparing information for you and you can basically, uh, then help, uh, deciding it. So from a meta perspective, how we, how we look at it.

Speaker A: Got it. No, it's fascinating, Gar. I mean, I mean, gosh, I have another 50 questions about that, but maybe I'll just ask, uh, one follow up, which is when you look at, you know, that question presupposed, that stuff does go wrong, right? When I asked you about where does liability land when something goes wrong, right. A mistake is made. And I wonder, are you looking? Are you tracking? You know, I don't want to call it quality, precision, accuracy. As these agents roll into production and move to Scale deployment. Are there metrics now that you pay close attention to that, you know, regarding accuracy and precision and quality of output and you know, integrity of this process execution? Right. Um, is that something that you're looking at as a product stability and product maturation perspective?

Speaker B: Ah, absolutely. As it is, I mean, I would say this is almost the, the main task in this new world engineers are doing. So we, you know, we call it the eval. So how do you create evals to understand is the agent doing what the agent is supposed to do? And we are testing this, of course, when we develop, uh, these agents. So we do this internal testing, but this is not with customer data. It's just like, you know, at least based on synthetic data or at least the test cases. We are saying we are in a good shape to let this basically now go into customer systems. But of course, again, customer systems means, you know, we have this private previews as a different stages. Uh, but of course also in customer systems we uh, provide this E wires for customers because as you know, every ERP system is a little bit different. So they can extend systems, they can configure systems. Uh, so it's important to understand are the agents doing the right things in the context of a customer system, in the context of the customer data with the extensibilities. Um, and therefore these E wires are even more important, uh, at customer side. And to your point, Mark, of course everything has a trace. So like in finance, if a human is doing transactions, there's an audibility trace and every agent has an id. So there is no somebody's messing up your system and you don't know who was doing it. So there is a trace. And you could say, okay, agent ID did this posting. But this agent ID is also connected to a user, uh, so it can be a right agent running around there. So when I'm, for example, I'm the head of the finance department, I want to have certain agent processes running there attached to me and of course I'm in charge that you think are, uh, also going well.

Speaker A: Interesting, interesting. And it's also trying to think of an example you probably could better than I could. But in some cases, would it be fair to say that some agents, um, say degrade or maybe there's there are leading indicators of imminent failure or imminent mistake or imminent wrong move. Right. Or is it more binary where there is no way to predict when an agent might screw up? Right. You know, maybe, maybe it depends on process or the agent or the human overseer. But I just wonder Are there ways to anticipate and mitigate, um, the, you know, failures and failure modes, as in agentic workflows, as opposed to just reacting to failures, so to speak?

Speaker B: I would say from, from our experience, the more likely cases, you know, you have, let's use expense management as one of our agent scenarios. You feel you have a good accuracy understanding, you know, invoices, hotel bills, you know, in America you have to itemize these things. Super complicated. Yeah. Um, we have a pretty good, at least, um, view on the test data we're doing. So we're giving this then to our customers. And I would say you would see two scenarios. Maybe the quality decreases. Um, and this could be, for example, oh, these invoices are in different languages we haven't really tested. Or these are maybe invoices, uh, of a different complexity. Because, I mean, of course we don't know all the invoices in the whole world. So. But I think this is also a learning mechanism. As we talk about maybe iq. I know if you heard about it, we have work iq, which is kind of representing the network of who is communicating with whom. What are the documents involved? And we see a similar network on the business process side. You have business objects, how they interact, how the process goes. And we want to basically have this iq. It's this intelligence how things are connected, but they also want to learn, uh, when I mean learning, I mean the customer wants to learn how their company works. So if they do, for example, expense management, maybe the agent failed and asked in human. Whatever, I don't know what to do with this one. Or I cannot assign it to the right cost item. So the human helps, but we basically understanding the trace. So basically the customer can use this the next time. Oh, now I understood it because we have done it. So there's this kind of improvement over time with the. But I think humans in loop are very important to at least memorize and build up this intelligence also for the agents to understand, oh, this is how this company works. And therefore, I know now, maybe the first time it took me 20 different iterations to find the right path, which of course was expensive because it creates tokens. But now, basically, I have done this before. Now I can go to the straight line. So we have these examples. There's many ways how you can drive from Seattle to New York. Um, you can go via Mexico if you want, but maybe you would like to go to i90 straight, you know, and. And that seems to seem like iq. It can help you to find the most efficient one

Speaker A: that's great. Garg, you mentioned tokens. It's probably a good segue. Move on to the next topic because we can keep going on agents for a long time. M. It's fascinating work that, that you're doing, but you think about the commercial side a little bit now. Um, tokenomics as new words coming off the, the assembly line here. But I mean as, AS sort of SaaS models and SaaS monetization models and subscription models have historically been seat based and user based. Obviously there's a lot of movement in the market with ISVs and ERPs changing the basis for um, consumption to things like, you know, assets or outcomes or other, you know, kind of drivers of usage, um, as, as the actors become less human or seat based, so to speak. So I'm curious, um, where you are, where the team is at at Microsoft with regards to these commercial frameworks, um, where you think you are now, where you're headed. What is the market really doing in your customer base? Or is this a demand? Um, is this something that you're pushing as opposed to customers asking for? Tell me a little bit about how you're, how you're thinking about this commercial evolution that's happening.

Speaker B: It's a fascinating one. And sometimes, you know, if you have been in this industry a little bit longer, um, there's the saying I think history is never repeating itself, but it often rhymes. Yeah, because if you go back to a little bit of broad observation, I think every impactful technology shift was accompanied by a change in the underlying business model. When the industry for example moved from on premise software to the cloud, the business model also moved from this perpetual license revenue to recurring subscription revenue. But I remember in the early, early beginnings every customer said I will never pay subscription. I mean, what the hell make this. I mean here we are. So I think something similar is happening when we move more to the agentic AI solutions. I think over time the business model will increasingly shift towards this consumption based pricing model. And I think the reason is pretty simple. In the cloud era, if you really think about it, you create software, you run it in the cloud, adding another customer to the solution. I mean almost came at zero marginal costs. I mean it's not completely true, but it was a very small amount of running 10 customers, 100 customers, thousand customers. It was the great scale of SaaS software by every investor of course liked it because it was this predictable model of increasing revenue and basically your costs are uh, not growing as fast. But now we're in a completely different economics because with agentic AI every user you are adding interacts and consumes, computes and tokens which creates costs. So the scaling is a very different thing. So therefore I would say consumption, at least based on the current dynamics is absolutely necessary. Um, I think agents, they are talking a lot of this, you know, we are taking off this undecided work of people so that people have more focus on what, you know, basically move to business. So if I, you know, if I spend all day chasing last month's invoices to close a quarter, you know, I'm not really thinking about how we will beat the next quarter's targets. Uh, so I, I think the benefits that the agents are bringing, it's just going to be, you know, these productivity gains, this decision velocity and just a much more agile business. So I, as I feel we are getting there. Um, the conversation with customers are still, it's just not, you know, the mental budgeting process doesn't really fit this consumption based model as I fully understand that. You know, a CFO says I need to know my budget for this month. It cannot be uh, whatever 10,000 and maybe it's 20,000. So I think there is again the same in the early days of uh, the subscription, uh, billing when we move to the cloud. And if you think back, I mean, uh, how long did it take for recurring subscription models to become the dominant one? I think it was many, many years. I don't think it's like switching overnight but we see more and more um, customers maybe dealing with it and also finding the value.

Speaker A: Right. Yeah, it is fascinating. It's kind of like you said, we can learn a little bit from the past and uh, even as many of you, many things happening now are unprecedented, there's still some instructive examples. And I think it's always a matter of like you said, it's like taking, taking friction out of a rapidly modernizing landscape where you know, various things don't fit together the way they used to. Right. In terms of budgeting and in terms of um, thinking about the predictability of your spend. Right. As like you just said, as a, as a finance leader and you know, kind of requires an organizational agreement to take this journey, which is not a straight road oftentimes it's a windy road to say, okay, we're moving to a future state of how we plan for, consume and leverage enterprise software. Um, and we kind of have a map, but it's not perfect. And I think that's harder for some companies to digest and accept than others in my experience. But I'm curious, how are your customers traveling with you on this journey? Are they aligned sort of mentally and strategically in terms of what commercial models are going to make sense for both parties? Are you finding conflict or how is this manifesting in your customer relationships?

Speaker B: I think there's always two categories of customers. There's the early adopters, uh, who want to move fast. They see this, maybe it's a disruption or I see this as an opportunity to maybe gain competitive advantage. I would say they are moving with us very fast. And I mean these are the right ones to of course experiment because the one thing, and I mean, let's be honest, nobody knows how the future will play out there. So everybody can of course have opinions. But I mean, I assume your opinion is as good as my opinion. So we have to learn fast and iterate and make the right conclusions. And yes, there's of course other industries which are maybe not really saying, hey, um, I don't have to be the first one, let the other ones figure it out and we learn from it. But I would say from a conversation perspective at least when we have our customer meeting. And so it's not like that they're pushing back and saying consumption doesn't make any sense because as I explained, it kind of makes sense because now every user is consuming something else. It's very similar. Like in engineering now every engineer is consuming tokens to create software. So there's, there's a very consumption based metric to it. So that's how you can capture the value. But I think what customers really push for is transparency. How do they know what's consumed? Of course the ultimate goal would be could you connect this to value? I don't have the answer there. As I just know already, 10, 15 years ago there was, I think a time that we tried to do value based pricing. But I think these conversations, at the end it was a disaster because I mean at the end you could not even agree how to measure the value. And then who is really driving this value? Was it not a software? Was it the customer? Was it the change process? Uh, I think this might be a little bit tricky, but of course value based pricing with consumption would be maybe the best one, but I think that's maybe the most tricky one. And we have this compromise where we are saying, which is maybe a good transition, you have a user model as I think everybody agrees we maybe not increasing our user base as maybe you have your finance users, but your company is growing and maybe you can do more and more with the existing user base. So I think there's an understanding that the user model itself is maybe not the great way of having a fair value capture for both sides. So we are trying to say, hey, you have a user and every user gets for example a certain, the amount of tokens they can consume as at least this is predictable. And then if you're going above these limits, you maybe would buy another package a little bit like Internet providers are doing. They're kind of saying it's unlimited bandwidth, but if you read the fine print, they're saying if you go amount to certain, then hey, let's talk about it because you may be running, uh, whatever, uh, a private hosting business. It's not anymore your own consumption.

Speaker A: Yeah, it is. I mean it's fast. I mean it really is. Um, I mean it's, it's interesting because obviously what your customers are trying to do, they're paid to do figure out the most, the fairest ROI lens. Right? How do, how do we, how do we look at return on our investment? Right? I mean it's the, that the rules of ROI have not been repealed, right. We're still looking for return. And so the question becomes, well, where is that return manifesting and what form of value and is it direct or indirect? And how do I attribute the driver of the right driver of that value and you know, fairly. And, and I guess, you know, and, and how do I protect my interest so that I'm not um, you know, moving backwards from where we were on a subscription seat based model, right. So that we're not somehow getting less value than we were before unknowingly. Right. So I think there's a lot of, a lot of um, well intentioned inquiry going on. Now my estimation, uh, you know, not just your customers, but the world of enterprise tech buyers that are trying to figure out how do you check all these boxes against this rapidly moving landscape like you said, where it's very difficult to predict what's actually going to be true in 12 to 18 to 24 months. Right. So lots, I mean, gosh, uh, you guys are leading the way in so many areas. Um, you know, and the customer, maybe I'll just. Yeah, so the dovetail into the next question. I mean the customer is such a driving force for what you guys are doing and what the best companies that are building solutions are doing is that they're staying very, very tight with key customers. And, and we hear, we hear about the successes, right? We hear about in particular Microsoft customers that are out there and publicly talking like Lifetime products and Coca Cola. They're using actual agents, you know, in areas like account reconciliation and areas like product change management. And they're attesting and testifying to, um, process improvements and, um, cycle time reductions. And they're actually giving words to some of these outcomes, which is fantastic because it's very powerful. It's helpful to you. Right. Um, but I'm curious about before those successes happen and before those customers are willing to go on record to say we got this return on our investment in this agent. There's got to be a lot of learnings that happened before that. Right. And I'm wondering if maybe you have an example or two of customer outcomes that maybe were in the earlier days with agentic deployment, maybe at smaller scales, um, where there were failures or there were misses that really instructed you and, and elevated the game and got you to that success. Right. Um, you know, we kind of get the. Get into the sausage making a little bit. What, what maybe, maybe some examples that come to mind.

Speaker B: I mean, there are so many examples because basically you learn something every week and they always say you learn m even more from failures maybe than from success. These are the really interesting ones, as long as you don't take them too close to your heart. Um, but the one big thing which was surprising, working with customers is, you know, how, of course, quickly this technology evolved, but also how much or how fast, basically customers moved and adapted and raised their expectations. And at least in my time now in this market, this is not true. For every technology disruption, mostly you have to almost push it. Like, come on, it's a big case, let's do it. Uh, um, but not be fair. I felt that it was a huge pull from customers and we launched our first agents once. You also mentioned, I think it's almost like two years ago and it was by design because we believed in this exponential technology shift, there is a first mover advantage. And, um, we built the agents again based on the knowledge we had. Two years. Of course, now in hindsight, sometimes even funny if you look at it, but basically we tried to do it, uh, with the tools we had at this time. So when we delivered these first agents and we went to customers, I mean, they loved it. And I said, that's a great example. As you called out the procurement agents, the account recognition agents, but then the customer said, hey, I have another hundred agents ideas. And we were thinking it took us a real good time to deliver these two or three agents. We are not building another hundred agents. Somehow something didn't really add up in these conversations. And I think this was the first really fundamental shift we had to do, well, as you maybe call it the messy middle, that we had to rethink this model first architecture. So what does it really mean? Model first? We're coming from this code first as you program something in code and then you give it to customers. Now we're in this. How can a model basically solve a customer problem? And, and in this first generation of agents, it was a very tightly coupled combination. It was like an application code with prompts, with large language models, everything kind of meshed together. Um, and these foundational models, I think every couple of months they're improving, there's other models coming, they're getting much better. And of course customers were expecting always running the latest models to get basically the best, uh, outcomes. Um, and with our initial architecture updating agents, it took us a huge effort to do that. And um, it reminds me a little bit, in the time when we moved to the cloud, the big thing we had to solve is how do you ensure every one of our customers runs on the same code base. There is no upgrades anymore, it's continuously delivery. You're always on the same innovation train. And now in the age of AI, we have a similar question. But the question is now how do you ensure every customer can always take advantage of the latest models or uh, the models of their choice? It can also be a choice by cost. Maybe you don't need the speed or the accuracy. And I think when we introduce Cowork now I assume a little bit, the context is known. I think all the puzzle pieces came together and it was much clearer. What does it really mean agentic erp, if I, you know, break it down as, instead of, you know, waiting for Microsoft as you know, the classic conversation we had in the past, oh, have a feature request and they talk to us and then we say in six or nine months maybe we can deliver it. I think there's now a different way and you don't have to wait for the agents. Basically what, how we changed our thinking is through the MCP server. We are basically exposing all the capabilities an ERP system has. So anything a user can do in our systems, this is not now erp, it can be CM any system we have there fabric now an agent can do. What we achieved with this step was that we are kind of now agnostic to the orchestration layer on top. So we have Copilot, for example, if you want to interact with the system. It's more like we are talking. I'm waiting for your response. And I respond to this again as it's more like a chat experience. But there's many other more complex processes. In a company where we have cowork, basically you're more like delegating a business process, and then the agent runs maybe in the background and you're not watching the agent, it's coming back. Um, and then either it needs more information or it comes back with a proposal. So basically what it means now is we don't have to deliver all these agent scenarios. We basically have changed. Users are interacting via Copilot or Cowork with our systems. And that's basically how you can define what I want to do. As again, in the example of expense management, I don't have to create now an agent to say, take a of piece, picture of my invoice or emails coming in, you know, itemize it, post my system. You can do this in, in your cowork app or Copilot. I'm just saying, oh, here's a picture of my restaurant receipt. Itemize it, post it to my dynamic system. And I think there's much more flexibility now. And, and I feel customers really like this independence. So they don't want to be, you know, boxed in again. And every time they want to change something or they want to enhance something, you know, they have to pay a lot of money to get it done. So I feel there is much, uh, more democratization how you can use your assets you have via this MCP server and this clear separation of the orchestration layers above it,

Speaker A: uh, it's. It's good. Oh, gosh, it's fascinating. It's great stuff, Georg. I mean, I think you're. It's a great time to be, uh, in the role you're in. I, I guess I'll say that. I think, I think you guys are leading the way in a lot of. A lot of dimensions and, and in a very uncertain future, which is tough to do, um, but also fun, also exciting, at least from where I.

Speaker B: Many, many CFOs, specifically, you know. So if you have been in this market for a while, as everybody's kind of agreeing, it's maybe the most exciting time for erp. Yeah, I mean, there was always the, the promise that ERP will help you to increase your productivity. As again, at the end, it's about resource management. I would say if you would talk to customers, then maybe it's like flipping a coin. 50%, but say, yeah, uh, it kind of helped me. 50%, but say, uh, it was more expensive than planned. It took More time. It didn't really deliver on the value, but at least there's a conviction now that these new technologies can really help. Because the thing in ERP is that there's still so many manual processes and uh, these kind of repetitive tasks, specifically in finance, I mean they're just really ripe for these kind of innovations. And therefore I think people really see now it's the time maybe to really uh, deliver on the promise of ERP and really help companies to run much more efficient.

Speaker A: Yeah, you mentioned finance, Georg. I've been involved in quite a lot of really, um, interesting discussions about the changing office of the CFO and, and how the CFO and his, his or her remit, um, is changing for the, for the better for the most part. And I'm kind of transitioning here a little bit to our speed round as we, as we move towards adjourning for the day. But when you think about finance, you know there's, there's zero room for, for error in many cases. I mean there are, there are people's jobs on the line for publicly traded companies. Right, as, as we all know. And so um, and also to your point about readiness, I mean there are accounting functions and massive time consuming, repeated processes like you know, financial close and things like that and um, that are just ripe for this type of innovation. And I'm curious, when you think about the various finance roles and job functions, um, we're already seeing some of those jobs evolve, disappear, get replaced by new jobs or new functions. Um, and I'm curious, from your seat, you're interfacing with a lot of these finance organizations and what jobs are you seeing in finance that are fading and which ones do you think are emerging?

Speaker B: I'd say a lot of conversations exactly around the questions basically you have raised. And specifically CFOs are super interested in understanding a little bit where the buck is going. I personally feel job titles are not disappearing overnight. Um, I think it's more like the work inside those roles is changing. Um, so you mentioned take this responsibility for a month and close, which is a very tedious process and you're running through it every month. And we were chasing this idea already many, many years ago. How could we move to a continuous close system process, which is the dream of many CFOs every day? On the tip of my finger I have a closed system because the longer the close process basically takes, the longer you're kind of maneuvering in the dark because you can only make decisions for the future if you kind of huge incentive um, and you know, instead of now spending days in manually reconciling ledgers, which I think this is one of the main activities they're doing and trying to chase, you know, these variances, you know, why is there a difference? Trying to explain it. You know, they're more like overseeing this continuous close process and more like reviewing the exceptions the agents are surfacing as agents can try to of course reconcile. Uh, there's a lot of rules, you know, you can apply there and at the end, you know, I mean maybe 80% of them you can resolve and then 20%, maybe the humans have to do. But this would be already a huge increase or reduction of time for the close processor. So if I had to invent a new job title, I think would be more like it's a finance agent manager or finance agent orchestrator. Someone who is supervising, govern and continuously improves. Basically a team of agents working across finance systems helping for example, uh, the close process. So I think what's becoming obsolete isn't the finance professional. It's more like this repetitive works that gets automated so, so people can hopefully spend more time on judgments on strategy and business impact and specifically finance. I think the one drain every CFO has is every time we grow our revenue. That's awesome. But I also have to grow the manual efforts to close the Book Studio and they would like to break out from. Can I decouple my revenue growth from my cost growth? At least let it grow slower would be already awesome.

Speaker A: Yeah, that's right. So you're expanding revenue and profit right at the same time.

Speaker B: Yeah, yeah, Uh, I guess

Speaker A: that's great. No, there's a lot there. But again, we're, we're gonna, we're gonna keep driving to the, to the end. You know, I'll, I'll, I'll ask you, um, a fill in the blank question here. Gay. Org. The biggest thing people, biggest thing people get wrong about AI and the enterprise is blank.

Speaker B: The biggest thing people get wrong about the I and enterprise, I think is that it's, that they think it's a technology problem. Because I think the technology is, it's already there. It's, it's working. Of course it's not working for everything, but I mean there's many use cases where it's working extremely well. But I think the real challenge is more like a, uh, is more like a leadership challenge, change management challenge or. So how, how do you adapt your organization the way you're working to this new, uh, technology advancements?

Speaker A: Agreed. Agreed. So back to maybe, uh, Something you mentioned earlier about not going into a room thinking you're the smartest one there. What would your team say about how gay org runs a meeting?

Speaker B: I guess you have, you have to ask them again.

Speaker A: That's true. That's a fair question.

Speaker B: I mean, I, at least I hope, uh, I hope they would say maybe three things as a. Yeah, the first one, I come prepared. Um, at least I, I like to be prepared.

Speaker A: Yeah.

Speaker B: The second one, I think as I mentioned, uh, a little bit the beginning, everybody has a voice. I don't think the best ideas come from the most senior people. Specifically now in this technology disruption, it's almost like upside down because experience is sometimes, you know, uh, a good and a bad thing. Yeah. And the third one is, you know, I always like to leave or at least conclude with decisions. Um, as I don't like meetings for the, you know, creating another meeting as I know it's sometimes a little more complicated, but, um, using the time most efficient.

Speaker A: Good. No, that's good. It's good stuff. Back again, two more questions and then we'll let you go. Garg. One, one is, um, hearkening back to your earlier days. We talked about earlier when you started your career and you have perspectives now as we do, as we've spent years in this market where big players like SAP and Microsoft, where now you've spent a good degree of time. Um, and as you've seen those competitive dynamics now evolve and like I said, constantly and very rapidly evolving as partnerships are struck and like you said, where companies are deciding to be agnostic about areas where they used to be, you know, fiercely protective as an example. And these dynamics are changing a lot. And I'm just curious your point of view on competitive dynamics right now. Um, maybe pick SAP because you've been there and can speak to the time is, you know, what, how, how are you viewing competition these days? And again, this doesn't have to be a lengthy answer as we, as we're concluding here, but just maybe a couple quick thoughts on do you think a lot about competitors or are you just really focused on, um, customer intimacy? Obviously it's not a binary. The answer is both. But you know, how are you thinking about competition these days and how is it driving your decision making?

Speaker B: I think at least a lot about competition also, you know, understanding who is your competitor. As I think also there is, uh, maybe a shift. Um, and of course, I mean since I worked for SAP a long time, there's also a lot time to reflect and think about. And I, I mean I have a lot of respect for both companies. And it's also interesting, I think the, the companies are optimizing for very different problems to solve, which of course also took me a while, you know, when I joined Microsoft as SAP has of course a decade of expertise in running the world's most complex business processes. But Microsoft has a really extraordinary ability to bring these new technologies to hundreds of millions of users. And these are very different starting points. Uh, and if you come from a different world, sometimes you're struggling to put these things together, um, because it creates all the different priorities. So I think, at least from my perspective, what I'm always trying to do is how do you combine the good things of both sides? Uh, um, what's the customer benefit when you try to understand both perspectives? So there's complex business processes, but I think all these productivity tools Microsoft is providing as well, I think these things are merging much more together and I think the lines are not so clear anymore and to competition as I. There's of course I call it traditional competition specifically now in the business process side. But if you look at, you know, so the native AI first companies like OpenAI Anthropic, it's also very interesting how they see the commercial business, how they try to enter, um, also at least a lot to learn from how they do it without this, um, 20 years of experience. As I said, sometimes, uh, it's a good thing, but sometimes also maybe keeps you too much in your own little books, how you're thinking.

Speaker A: Yeah, that's right, that's right. No, I agree, agree with that. Well, last question, G. This is, um, if, if Mr. Sachin Adela were sitting here with us, which of course he's not, and, and you were able to, um, or you and I, let's just say were able to ask him one question, any question, what would you suggest we should ask and what do you think his response would be through the lens of helping Microsoft customers and maybe prospects get a better grasp of what do they stand to gain from working with Microsoft in the ERP space.

Speaker B: If Satya would be sitting here, so my first reaction would be, wow, Mark, you really upgraded your guest list.

Speaker A: That's true.

Speaker B: Um, on the question, as I, in the context now of Dynamics, and I'm trying now to assemble a little bit also what Satya was saying specifically now Build conference was very interesting. It's a little bit how he would describe this. AI agents unbundling traditional SaaS software or SAS tech. So as dynamics basically is evolving into this AI first business platform. What does he believe will become this source of differentiation as we just talked about competition as well as the thing customers simply cannot replace. And how would this basically change the way we think about our ERP products, for example, or CM products today? Um, this could be one question at least I was asking myself a couple of times if I'm trying so to put all these puzzle pieces together from Satya's messages. He's doing publicly.

Speaker A: Great. That's great. Yeah. Obviously he's got a lot on his mind, for sure. But that's, that's, uh, that's very astute. Well, good. Well, Georg, this has been tremendous, you know, and I, I think we could. We, uh, could go on for hours, but. But you don't have them. I don't have them. So we're going to adjourn for now. And m. I want to thank you genuinely, um, Garrick, for taking the time, uh, always a pleasure to chat with you and learn about where you're headed and, um, what you're learning. Because I think you're right. We're all learning at this stage and we all have different, um, perspectives on, on, um, um, what we should be planning for in the near future and maybe even medium term future. Um, so thank you for that. I know everyone else benefited. We've got lots of people in this community that are, that are trying, um, to navigate, obviously finance workflow, but also supply chain and CRM and marketing and, um, you know, production and obviously, uh, HR and accounting and things. So all very much in scope for erp. So thank you for your time today. Thank you everyone for listening. Appreciate your ongoing attendance, uh, and participation in our community. We will catch you next time on the next episode of the Enterprise Edge. Until then, take care and, uh, we'll talk to you soon. Thanks, everybody.

Speaker B: Thank you. Mark.