Latest / The Adults In The AI Room / Earley AI Podcast – Episode 87: AI-Enabled Enterprise Data Migration with Dominik Wittenbeck

Transcript

Earley AI Podcast – Episode 87: AI-Enabled Enterprise Data Migration with Dominik Wittenbeck · The Adults In The AI Room

• 0:01
  SPEAKER_00: This is the Early AI Podcast hosted by Seth Early.
• 0:04
  SPEAKER_00: Join us as we delve deep into the passions, expertise, and experiences of thought leaders and practitioners to talk about what's possible with artificial intelligence.
• 0:14
  SPEAKER_00: The Early AI Podcast is sponsored by Early Information Science, your digital transformation journey with design and deployment of innovative technology solutions, as well as Vector.
• 0:26
  SPEAKER_00: Now enjoy the show.
• 0:28
  SPEAKER_02: Good afternoon, good evening, good morning, whatever your time zone is.
• 0:32
  SPEAKER_02: Welcome to the Early AI Podcast.
• 0:34
  SPEAKER_02: And in our podcast series, we basically talk about what organizations are doing to be successful with AI, what the challenges are to that success, how to get from proof of concept to production, how to deal with data issues, how to deal with agent issues, security, compliance, all of those types of things.
• 0:52
  SPEAKER_02: Today we're going to talk about something that's very high stakes for organizations.
• 0:56
  SPEAKER_02: And that is really, you know, when you're doing enterprise data migration, how does that impact the organization when it comes to AI initiatives?
• 1:03
  SPEAKER_02: When you're doing a merger, when you're doing a divestiture, an SAP platform upgrade, you know, the data has to move.
• 1:10
  SPEAKER_02: And so if something goes wrong during that process, there's certainly significant impact to the business and it can be very consequential.
• 1:18
  SPEAKER_02: And of course, when we are trying to leverage AI in all of these initiatives, which is what we're doing now, we introduce new factors, new risks, new compliance issues, new challenges that didn't exist necessarily in prior iterations of this.
• 1:32
  SPEAKER_02: We've always used AI to do things like data cleansing or use machine learning, shall I say?
• 1:38
  SPEAKER_02: You know, we used to call it things like text analytics and data analytics and data architecture.
• 1:42
  SPEAKER_02: And now everything is AI, but it's still some of the fundamental principles that we've used before the advent of generative AI, which is the recent iteration that people are so excited about.
• 1:54
  SPEAKER_02: So there's very little room for error when we're doing things like migrations.
• 1:58
  SPEAKER_02: There's one anecdote, and I'm not sure if you gave us the uh this, uh, Dominic, but$5 million per hour when uh a system was down or when an organization was down, an enterprise critical system was down.
• 2:09
  SPEAKER_02: So my guest today is Dominic uh Wittenbach.
• 2:11
  SPEAKER_02: He's group CTO of SP Group, which is a 1600-person global consulting firm that has spent 30 software and solutions firm, that has spent 30 years doing SAP-centric data migration.
• 2:23
  SPEAKER_02: Dominic has been with the company for 25 uh of those years, and uh he spent most of that time doing something that uh will resonate with our listeners talking about um what's inside the heads of experienced consultants, codifying that into structured, reusable knowledge.
• 2:39
  SPEAKER_02: SNP now has what they call a transformational repository covering millions of data points across the SAP data model, plus guided procedures with 1,500 discrete steps for running a migration.
• 2:52
  SPEAKER_02: This is 15 years of formalized enterprise expertise.
• 2:55
  SPEAKER_02: And now they're building agent AI systems, agent-based AI systems to work on top of all that knowledge repository.
• 3:02
  SPEAKER_02: And this is what we're finding is that organizations that are doing extremely well with AI are using those repositories, are using that uh institutional knowledge and using that to drive many of the AI uh initiatives.
• 3:14
  SPEAKER_02: So, Romnik, welcome to the show.
• 3:17
  SPEAKER_02: Thank you for being with us today.
• 3:18
  SPEAKER_02: Thank you for having me, Seth.
• 3:19
  SPEAKER_02: Thank you for having me.
• 3:20
  SPEAKER_02: So um so uh we are in uh uh the northeast and we're getting more and more snow here.
• 3:25
  SPEAKER_02: How is it in uh your neck of the woods?
• 3:27
  SPEAKER_02: You're in you're in Germany, where exactly are you?
• 3:30
  SPEAKER_01: In Heidelberg, and actually I can say we had kind of a winner this year for like, I don't know, decades probably.
• 3:37
  SPEAKER_01: My little ones have been seeing snow for the first time for real.
• 3:40
  SPEAKER_01: Wow.
• 3:40
  SPEAKER_01: Um but I mean it's it's it's starting to get spring here already.
• 3:44
  SPEAKER_01: Yeah, we had like the first warm couple days basically.
• 3:46
  SPEAKER_01: Uh nights are not cool so much anymore.
• 3:48
  SPEAKER_01: So I think we've we're hoping for spring weather.
• 3:51
  SPEAKER_02: So we'd like to start our get start off by asking people about misconceptions.
• 3:55
  SPEAKER_02: So you operate in a domain where the consequences are very significant.
• 3:59
  SPEAKER_02: So when you talk to executives about applying AI in your space and data migration and doing these redeployments, replatforming, what are the biggest misunderstandings that you come across?
• 4:09
  SPEAKER_01: I think mostly it is like AI will solve basically everything.
• 4:13
  SPEAKER_01: I mean, the field that we're in, and you already like expanded on this.
• 4:16
  SPEAKER_01: This is born out of an expertise out of like two decades plus, basically, where I think I spent most of my professional career with extracting the knowledge from consultants and trying to formulate it in a way that actually fits into software, basically.
• 4:32
  SPEAKER_01: And now when I think the biggest misconception is that AI is going to be the silver bullet that basically heals all out of like thin air.
• 4:40
  SPEAKER_01: Um, but actually, ultimately, what it needs, I mean, everybody is like aware of that, that um, in order to train AI, you need data.
• 4:48
  SPEAKER_01: Nobody would question that, even, right?
• 4:50
  SPEAKER_01: But in order to run complicated and like life-depending systems, but systems which basically have a critical impact on your business, they need enough grounding.
• 4:58
  SPEAKER_01: They need to make sure that whatever an agent runs, it really runs on a on a foundation that is probably not out there in common knowledge.
• 5:05
  SPEAKER_01: So you actually need to feed the knowledge into the agentic system that you're building on top of in order really to make sure that you get deterministic results in the end.
• 5:15
  SPEAKER_01: And I think the misconception is that, hey, this all becomes very easy, and making it basically a I can save time, I can save cost type of problem, where in reality, what you really want to make sure when you're getting yourselves in the data migration is that you're doing it correctly, that you're doing it efficiently, that you're doing it safe and sound and audible in the end.
• 5:36
  SPEAKER_01: Because I mean, ultimately your system is going to get audited, and whatever stuff has been done in there incorrectly, it will surface at some point.
• 5:43
  SPEAKER_01: And you have to like make sure that this doesn't happen at basically all costs.
• 5:46
  SPEAKER_02: Yeah, that's a great point in the fact that it is about risk mitigation, it's about doing it correctly, it's not necessarily about doing it as cheaply as possible because sometimes uh cheap is is great quite costly, and sometimes uh slowing down speeds things up.
• 6:02
  SPEAKER_02: So we have to be uh thoughtful about this.
• 6:04
  SPEAKER_02: One of the things that we uh talked a little bit about was that you know, data migration requires that you have teams of specialists who understand the sources and the destination systems, and that you know, an individual there's too much complexity to hold in one person's head, right?
• 6:22
  SPEAKER_02: And that's where you know you have a lot of friction.
• 6:24
  SPEAKER_02: We have different people that have different areas of specialization, but and many times that is difficult to scale.
• 6:30
  SPEAKER_02: So clearly, this is you know an expertise, a classic expertise bottleneck.
• 6:35
  SPEAKER_02: And how has that scaling constraint you know shaped your approach to AI?
• 6:40
  SPEAKER_02: Obviously, this is something where you know it's a it's kind of an ideal use case, but tell me how you've thought about it and tell me maybe what some of the the experiments look like and then how you formulated your your approach that has uh been working for you.
• 6:54
  SPEAKER_01: Yeah, I mean, you're exactly right.
• 6:56
  SPEAKER_01: I mean, migration is a team store basically.
• 6:58
  SPEAKER_01: If anything else, I mean you always say when you build a product or when you build anything, you want to have basically functional teams, right?
• 7:04
  SPEAKER_01: You want to have interdisciplinary teams that you mix people from the business with technical people to actually get the best outputs and to get the best results.
• 7:11
  SPEAKER_01: Now, in data migration, even you're basically operating, well, you're turning, you're modernizing a system um basically full-fledged all the way through all the processes you're going to touch, all the data you're going to touch.
• 7:22
  SPEAKER_01: Yeah, and you have to make sure that everything basically is perfect.
• 7:24
  SPEAKER_01: So you have to get a lot of people actually into the game, both from the customer and also, of course, from the um from the SI or from whomever does the data migration uh in this particular piece.
• 7:34
  SPEAKER_01: So, how does AI really help in this particular regard?
• 7:38
  SPEAKER_01: And to be honest, it starts even way before you actually hit any particular data transformation tooling, like you might think.
• 7:45
  SPEAKER_01: I'll be talking about like building data pipelines.
• 7:47
  SPEAKER_01: And yeah, believe me, I will talk about building data pipes and stuff like that.
• 7:50
  SPEAKER_01: But the problems or the challenges even start way before, right?
• 7:54
  SPEAKER_01: Because what you typically start with when you are um being presented with a data migration opportunity at a customer, they start with an RFP.
• 8:02
  SPEAKER_01: And it it shows like where do they actually want to go.
• 8:05
  SPEAKER_01: It doesn't necessarily state where do they actually come from.
• 8:09
  SPEAKER_01: So what we experience always, and that was even before AI was a thing, so for like 15 years or so, our approach to the subject always was to get as much transparency into the process of the S is situation so you can evaluate very precisely how to actually get to the to be situation, right?
• 8:29
  SPEAKER_01: So we were always running something like a scan on the source systems to really truthfully understand facts and figures about how the system was used, which modules were in place, which processes were run, which data was actually on these systems, to what degree was it on there, and so on and so forth.
• 8:44
  SPEAKER_01: This type of things.
• 8:45
  SPEAKER_01: So, with that said, what you need in any AI operations is foundational knowledge which is specific and not generic to the underlying problems.
• 8:55
  SPEAKER_01: So you need that data to understand and flesh out the way into the future.
• 9:00
  SPEAKER_01: So even at the very beginning of such a project, when you're fleshing out the solution, say like, okay, this is about like how we're going to move the data.
• 9:07
  SPEAKER_01: That is the decisions that you need to make because of the inconsistencies that see on the source.
• 9:12
  SPEAKER_01: Using AI to basically bring all this together, even in something just as condensed as a very distinct and concrete and built to the customer survey that makes sure that you're really hitting all the points and all the decision that needs to be made can be made as the earliest point in time, is something where we where AI basically already helps.
• 9:33
  SPEAKER_01: And that's not even downstream building it too much into a platform that actually does the data movement and actually does the data migration.
• 9:40
  SPEAKER_02: So maybe you could put a little bit finer point on this.
• 9:43
  SPEAKER_02: Um, so so this is a situation where you need to bring specialists together.
• 9:49
  SPEAKER_02: You have reference uh procedures and rule bases, rule bases and and methodologies.
• 9:55
  SPEAKER_02: So you're not necessarily just bolting on you know a virtual assistant to answer questions.
• 10:01
  SPEAKER_02: How are you actually using this to accelerate what the team is doing?
• 10:06
  SPEAKER_02: So to give me kind of a before and an after, right?
• 10:09
  SPEAKER_02: If you did a project a couple of years ago without any AI assistance with minimal compared to what we have today, what did that look like versus what it looks like today using some of the agentic approaches?
• 10:21
  SPEAKER_02: And again, how is it kind of um integrated into the workflow and the work processes?
• 10:26
  SPEAKER_02: Is that okay?
• 10:27
  SPEAKER_01: So let's let's let's virtually run through basically a project and basically see at what points right uh stuff has changed because of agency capabilities that basically came into play.
• 10:36
  SPEAKER_01: So if we basically pick it up of where we just left it off in the conversation, at basically the pre-sales phase at the solutioning phase, where you're trying to find out where actually you want to basically be and how that uh the thing is going to be laid out, one of the next things that typically is happened, you basically craft a project timeline.
• 10:53
  SPEAKER_01: You craft basically uh a project step and methodology in there.
• 10:57
  SPEAKER_01: You craft a blueprint basically about how your migration is going to be.
• 11:00
  SPEAKER_01: Of course, I mean crafting the blueprint itself, if you give it the proper template, which we formulated basically over years, is something which is much easier to do if you can basically just well slap it into AI and get a good proposal to begin with.
• 11:12
  SPEAKER_01: I'm always saying a good proposal to begin with, because ultimately you, as a human being, as the person who's the solution engineer or enterprise architect in that particular case, you will have to take responsibility and accountability for the end result.
• 11:25
  SPEAKER_01: So I always take stuff that the AI spits out that point as a proposal.
• 11:29
  SPEAKER_01: So you have that.
• 11:30
  SPEAKER_01: Basically, you were perfectly prepped.
• 11:33
  SPEAKER_01: What you next take basically is you take this thing and you actually run and build a data pipeline from this, from the actual, from the actual pieces.
• 11:40
  SPEAKER_01: So it will take you through the different steps.
• 11:42
  SPEAKER_01: And the platform that we have has something like guidance procedures in them that, as you said, it's like millions of steps, no, it's not millions of steps, but it's thousands of steps, which are interdependent basically.
• 11:52
  SPEAKER_01: But they're not necessary for all the pro for all of the uh different pieces.
• 11:56
  SPEAKER_01: They're basically like crafting out almost all the eventualities that might happen.
• 12:00
  SPEAKER_01: And now with a blueprint in there, we can dynamically switch on and off all the things.
• 12:04
  SPEAKER_01: So the pre-configuration of this scenario that will typically be done manually, now of course it's been sped up much and much in that regard.
• 12:12
  SPEAKER_01: Next, what you typically have to do is you have to craft an individual rule base.
• 12:16
  SPEAKER_01: So when you have the procedure right and what steps you need to take, typically what happens is like, okay, we need these and these objects basically, and we have templates for that as well, but then you need to like adjust them because maybe you're changing the chart of account in a certain degree, right?
• 12:29
  SPEAKER_01: Or you're reorganizing your profit centers along the way, or your cost center structure, or your organization in general while you're doing that.
• 12:36
  SPEAKER_01: And you would typically go down that route because if you are an IT leader, what you typically want to do is you will want to become the enabler of the business.
• 12:45
  SPEAKER_01: So while a typical system modernization at first hand might seem like, oh, let's get it over and done with and make it as cheap as possible.
• 12:53
  SPEAKER_01: If you're like smart about it, you can also say, like, okay, this is a chance of change anyhow.
• 12:58
  SPEAKER_01: Let's incorporate business benefits that streamline accounting downstream, right?
• 13:03
  SPEAKER_01: Or that enable better reporting capabilities, which ultimately lead to better planning experiences.
• 13:08
  SPEAKER_01: And if you want to take that, this will end up in rules that you're formulating, like how you want to reshape your organization.
• 13:15
  SPEAKER_01: Now, these rules typically would in former times have been encoded directly in the software.
• 13:21
  SPEAKER_01: So you would configure mapping lists and stuff like that, and flesh out rules in scripts and so on and so forth.
• 13:27
  SPEAKER_01: Now, much of the things can now be scaffolded, of course, with with the help of AI, right?
• 13:31
  SPEAKER_01: So still we're using the underlying system and the way of how it's been executed, but the rule generation, right?
• 13:37
  SPEAKER_01: The crafting of the pipeline, that's what actually gets accelerated.
• 13:41
  SPEAKER_01: Now, let's say you have all of that done already, you actually now run the pipeline.
• 13:46
  SPEAKER_01: And what you really want to make sure is that it runs exactly as it would have run traditionally, meaning deterministic.
• 13:53
  SPEAKER_01: So from whatever same inputs you give it, you get exactly the same outputs because that ultimately is a necessity for an auditable system that you want to have in the end.
• 14:04
  SPEAKER_01: Now, I'll stop right there because at the end of the process, when we get to testing and we get to validation of those pieces, there's a whole lot more that AI can actually do nowadays versus former times where much of the testing actually would have done it need to be done manually.
• 14:19
  SPEAKER_02: So, you know, what you're saying is even when you're doing data migration planning, you are still looking at process engineering.
• 14:27
  SPEAKER_02: You're still looking at doing some optimization.
• 14:29
  SPEAKER_02: Is that correct?
• 14:30
  SPEAKER_01: Well, most of the times customers are looking for these optimization.
• 14:34
  SPEAKER_01: And you, as the vendor for doing the data migration, actually has to realize them and make sure that they actually do happen, basically.
• 14:40
  SPEAKER_01: So the initial intent or the initial uh spark typically comes from a customer's well craving for hey, let's become an enabler of the business.
• 14:50
  SPEAKER_01: Let's, while we move this whole stuff all together, take the chance of the change that we're gonna uh go through anyhow, and don't do another project a year down the road, but inline this right away this time.
• 15:01
  SPEAKER_01: So the time to value that I'm providing as an IT head to my organization increases.
• 15:06
  SPEAKER_02: Yeah, and and I always talked about the fact that if you're moving, say, stuff from one system to another and you say, Oh, we'll clean it up later, right?
• 15:14
  SPEAKER_02: That's never gonna happen.
• 15:16
  SPEAKER_02: It's like moving from one house to another, and you have a an uh an attic full of boxes and junk and stuff.
• 15:21
  SPEAKER_02: And when you're moving it, you just move it to the new attic and you say, Oh, we'll clean it up later.
• 15:25
  SPEAKER_02: No.
• 15:26
  SPEAKER_02: You're not gonna clean it up when you're moving it, you're never gonna clean it up, right?
• 15:29
  SPEAKER_02: So, so this is kind of looking at that, uh, you know, that from that perspective, but you're also you're also kind of so what does this really mean to an organization?
• 15:38
  SPEAKER_02: And and give uh, you know, when when they're going about these these initiatives, and tell me more about is it simply a matter of saying, I mean, what's the takeaway here?
• 15:46
  SPEAKER_02: Is it simply a matter of saying, look for your vendors to use AI to improve the efficiency?
• 15:51
  SPEAKER_02: Everybody's looking for that, or are there things that they need to do in order to prepare for a vendor coming in and doing this migration and doing this replatforming that they need to think about in advance?
• 16:04
  SPEAKER_02: Like what's different?
• 16:05
  SPEAKER_02: I kind of want to know what's the different what's new, what's different, what's the takeaway for AI-enabled data migration and replatforming versus what they've done in the past?
• 16:15
  SPEAKER_02: And part of what you're saying is look at the bigger process, look for where you want to build efficiencies, look for the you know the efficiencies that AI can provide, both in terms of the migration, the data cleansing, but also in terms of the functionality that you're looking for.
• 16:29
  SPEAKER_02: What else?
• 16:30
  SPEAKER_02: Is there anything else that you would add to that uh list?
• 16:33
  SPEAKER_01: I mean, ultimately, of course, we would be happy if like systems would have been cleansed already because it would make our jobs so much easier, right?
• 16:39
  SPEAKER_01: As you as as you just like made the example, when when you move over, you have like once of a lifetime or once of a legislation period time of chance, actually as a CIO to now make things clean again.
• 16:50
  SPEAKER_01: I mean, once they will decay and go downstream or will go downhill from there anyhow, right?
• 16:55
  SPEAKER_01: But it's like a one in once-in-a-lifetime chance that you that you actually can take and you can basically piggyback right on these things.
• 17:02
  SPEAKER_01: But you have to also like carefully choose, right?
• 17:04
  SPEAKER_01: Right, because the more change you incorporate, I mean, data cleansing is not so much of a change, really.
• 17:09
  SPEAKER_01: It's like more of a housekeeping task that you can do alongside.
• 17:12
  SPEAKER_01: But more of a change that impacts the business later is of course a burden that the business has to take within this project.
• 17:18
  SPEAKER_01: What we always say, like, hey, cater to what your organization or cater the project and the scope that you put in the project to what your organization can actually digest.
• 17:28
  SPEAKER_01: To be honest, that doesn't have so much to do with AI ultimately, right?
• 17:32
  SPEAKER_01: Because it's much about like how can you manage people, how you can get them excited about the new system that well sits there in glory on the on the mountain uh top, right there in the distance, and it's all going to be better.
• 17:43
  SPEAKER_01: And I mean, you're probably exhausted by the time that you get there.
• 17:46
  SPEAKER_01: Maybe AI makes it a little bit less exhausting to ultimately go there.
• 17:50
  SPEAKER_01: But um the change management that you need to do, that people on board, of course, gets harder the more change you accumulate in one project.
• 18:00
  SPEAKER_01: And that's something that you would need to consider.
• 18:02
  SPEAKER_01: And for me to be honest, that's something which is independent of AI.
• 18:05
  SPEAKER_01: The only thing that really really matters and helps here really is that making it clear what the impact is through like interpretation of analytics, the uh recommendations or the the summaries that you can get out of results of analytics make it maybe a little bit more tangible to make this presentable so you can make a more informed choice and don't do naive choices basically along the way.
• 18:28
  SPEAKER_02: So you make a good observation.
• 18:30
  SPEAKER_02: There's always the you know, this the pace of change uh uh exceeding the ability to absorb change.
• 18:37
  SPEAKER_02: And that happens at multiple levels.
• 18:38
  SPEAKER_02: Right now, it's happening with developers, right?
• 18:40
  SPEAKER_02: The pace of change, the technology is happening uh faster than developers can can process and assimilate and apply.
• 18:47
  SPEAKER_02: Now, of course, the AI tools are helping with that because they can take out a lot of the um uh a lot of the uh heavy lifting around coding, but then that still presents overload of what the capabilities are.
• 18:58
  SPEAKER_02: So if you can build these capabilities more quickly, you're you're giving the organization more things to absorb, right?
• 19:04
  SPEAKER_02: And we're we're seeing that quite a bit.
• 19:06
  SPEAKER_02: Um and and then you know, the technology changes faster than business can change, the marketplaces change faster than the culture can change.
• 19:14
  SPEAKER_02: Uh so there's this trade-off between a recency currency and the ability to absorb that change.
• 19:20
  SPEAKER_02: And as you say, there's a lot of that having to do with socialization and uh and change management more than it is the tools themselves.
• 19:27
  SPEAKER_02: So again, when an organization is looking at one of these catalysts around, say, replatforming, what do you see driving the migrations these days?
• 19:36
  SPEAKER_02: Is it the acquisition, the divestiture, the new platforms, or is it looking at AI native applications, or is it supporting migration and cleansing for AI?
• 19:48
  SPEAKER_02: What are the big drivers and catalysts that you're seeing right now in the marketplace?
• 19:52
  SPEAKER_01: Okay, I mean, maybe the market and our own perception might be a little bit different right now.
• 19:56
  SPEAKER_01: The biggest that we get yeah, from from from our perspective.
• 19:59
  SPEAKER_01: So the the biggest push that we are getting is of course because of the wave to S4, basically.
• 20:05
  SPEAKER_01: But to be honest, I mean that's a wave which like started off like 10 years ago already.
• 20:09
  SPEAKER_01: So but I guess SAP that's the SAP S4, yeah, yeah.
• 20:13
  SPEAKER_01: For everybody who's not not familiar, I mean SAP is one of the biggest vendors like globally to supply ERP systems, basically.
• 20:20
  SPEAKER_01: And they recently, I mean like 10 years ago, announced basically a shift from their ECC to the S4 platform, which is undergoing basically ever since.
• 20:29
  SPEAKER_01: The reason for why this has been taking so long for organizations to digest and even to like start adopting at this point is they invested for already for 25 years into these old systems, optimizing them.
• 20:41
  SPEAKER_01: And one of the key features or one of the key selling points for SAP always was the high degree of customizations that you can actually imply into the system to make it really diverse, to make it really like specialized, to optimize for your own processes and for your own benefit.
• 20:55
  SPEAKER_01: And then, of course, if you basically have to restructure this, maybe even go into a standardized application set in the future, makes you think hard about hey, what can you actually let go and what what can you actually should be taking?
• 21:08
  SPEAKER_01: So I think modernization is one of the drivers right now, but it's fueled by two things or two views.
• 21:15
  SPEAKER_01: One is in SAP specifically, maintenance for the old systems is going to run out.
• 21:20
  SPEAKER_01: So eventually I'm not going to support it anymore.
• 21:22
  SPEAKER_01: I mean, that's something that if that hits you, you are in a bad position, right?
• 21:26
  SPEAKER_01: Because you're basically being beaten with this project.
• 21:29
  SPEAKER_01: You have to do it no matter what, and you cannot choose the pace anymore.
• 21:32
  SPEAKER_01: Then, of course, the other side of the coin is that hey, this new system that we're going into, it's highly AI enabled, right?
• 21:39
  SPEAKER_01: And if we get all this data in there, especially also a historical one and reshape it and basically bring it into a shape and a form that now the new AI use cases available in that system could readily use this.
• 21:51
  SPEAKER_01: That is going to be a bright future that really enables new business models, new business abilities that we basically have.
• 21:57
  SPEAKER_01: So, this I would say are the two most drivers that we see.
• 22:01
  SPEAKER_01: The merchant acquisitions happen all the time.
• 22:03
  SPEAKER_01: And yes, of course, they also increasingly happen in this point.
• 22:07
  SPEAKER_01: But I'd say the way of how you handle those is more of a well, it's more of a bread and butter business for us in the sense the modernization piece is really where it's peaking right now.
• 22:18
  SPEAKER_02: So one of the things that we talk about with uh customers whenever they're doing things is having the right semantic foundation, right?
• 22:26
  SPEAKER_02: The semantic architecture.
• 22:27
  SPEAKER_02: What does that mean?
• 22:28
  SPEAKER_02: That's starting with business concepts and it's translating those business concepts into lots of different types of elements and processes and applications and workflows and systems and tools.
• 22:40
  SPEAKER_02: And what's interesting is when you look at this from a data perspective, sometimes you miss certain context.
• 22:46
  SPEAKER_02: And I want to give you an example and get your thoughts to this.
• 22:49
  SPEAKER_02: If you think about data, you're talking about data an element in a data field, right?
• 22:54
  SPEAKER_02: You're talking about a very fine grained application of a business concept to something that is structured or could be unstructured, but it's looking at that level of here's a here's a field and here's a value in that field.
• 23:06
  SPEAKER_02: Now when we when I talk about this from a conceptual perspective, I use the example of an of a business term and the an example of a collaboration environment like SharePoint.
• 23:17
  SPEAKER_02: If you have a concept called methodology, right, which is very important to consulting business, you could have a site collection that's called methodology, right?
• 23:24
  SPEAKER_02: So your business term, your business concept is methodology and you name that site collection.
• 23:29
  SPEAKER_02: You could also and so that's one way of applying that element from your semantic architecture.
• 23:34
  SPEAKER_02: Now there's another level of granularity we could say well I'm going to call a site methodology.
• 23:39
  SPEAKER_02: So I could have a site collection that has lots of sites and it's all about different kinds of methodology or I can have a site that's about methodology.
• 23:45
  SPEAKER_02: I could also have a library within that site and I could call that methodology.
• 23:50
  SPEAKER_02: I could have a list within the library called methodology.
• 23:54
  SPEAKER_02: I can have a content type within that library called methodology.
• 23:58
  SPEAKER_02: I can have a metadata field within that content type called methodology or I can have a value within that control vocabulary list in that field called methodology.
• 24:09
  SPEAKER_02: If I go right to the data architecture in the master data piece I'm going right to that finest level of granularity looking at the data in that field yet there are many other mechanisms and many other contexts so sometimes if people say oh just add a term to the taxonomy it's not that simple right because it can because it means different things at various levels of different things at various levels.
• 24:31
  SPEAKER_02: Do you um in your uh uh uh projects do you begin with that domain model at all to talk about these different contexts or do you mostly because again the nature of the beast is you got to migrate this data or do you mostly focus on the data elements so tell me how you think about the semantic architecture infrastructure in an enterprise.
• 24:53
  SPEAKER_01: Okay so I would say you don't have to have the answer just let's noodle out in no no no no no no there is not the answer anyhow because stuff is moving all the time but I would have an answer which is at least true from our perspective right so if you are in an SAP to SAP strict data migration field you are a very lucky person actually because SAP being a German company you can imagine how they overengineered and perfectly fine-grained formulated out the data model upmost to like the most granular thing being very very distinct.
• 25:24
  SPEAKER_01: So in an SAP system you find very little generic and shapeable terms.
• 25:30
  SPEAKER_01: What you find instead is emerge and I mean like hundred thousand database tables.
• 25:36
  SPEAKER_01: And those hundred thousand database tables in that system and I kid you not they have like 250 fields on average and these fields on average typically have like multitudes of distinct values.
• 25:47
  SPEAKER_01: What this means is that the the metadata that you have available and the the way of how the semantics are manifested in the system itself is very strict.
• 25:57
  SPEAKER_01: And once you had the chance such as we did to basically chew in the system for 25 years which of course evolved over time but thankfully not so often in a in a in a destructive way but more in an additive way we gotta learn a lot about how that data model works.
• 26:15
  SPEAKER_01: So in that sense actually the terminology the methodology the taxonomy is pretty much set.
• 26:21
  SPEAKER_02: Now there's an other side of the coin because but but but when you say it's set but it's it's very dynamic right because new products new markets new solutions from a value perspective it actually is.
• 26:33
  SPEAKER_01: From a value perspective it actually is right because I mean of course customers do encode their semantics and their processes specifically into the customization.
• 26:42
  SPEAKER_01: I mean I I said earlier on that one of the big things about SAP was how you can like twist and tweak the system to actually satisfy your needs.
• 26:49
  SPEAKER_01: So of course gaining an understanding specifically at the very very beginning of the phase like what is actually in there what data objects do they really use how are they expressing and how are they relating to the to the to the actual processes is of course of imminent importance because you're going to migrate that you're going to touch that you're going to probably change that along the way they need to of course then finally end up in the data structures or in the semantics actually that uh that that that uh they envision to be like their true business in the future so a lot of what we do and I typically refer to this as value mapping versus structural mapping.
• 27:26
  SPEAKER_01: So the structure in SAP pretty fixed the value very very fluent and so the value mapping and like finding is the how do you map the terminology from point A to point B, from system A to system B, is one of the dominant exercises that you undergo when you are in the blueprinting phase with the customer.
• 27:43
  SPEAKER_01: You sit there with customers in workshops talking about this value meant this before I don't find this in the target system configuration.
• 27:50
  SPEAKER_01: I found this XYZ other values that might have the same meaning the semantic meaning is this really the case and also here of course AI can help a lot because evaluation mostly then happens on things which are not strictly bound to the data model.
• 28:03
  SPEAKER_01: They happen on the text level where the semantics basically are implicitly encoded and this is also where AI can actually help very much.
• 28:11
  SPEAKER_01: So another point or another flip of this to turn it even to more of an extreme is when people want to move into an SAP system they come from a non-SAP system.
• 28:22
  SPEAKER_01: Classical case of re-standardization I build a homegrown tool for some sort of process in my corporation that guy that built it back then or that girl that built it back then they're long gone they're retired and actually this imposes a big business risk.
• 28:36
  SPEAKER_01: So hey can we get rid of this system as we are moving into a new one?
• 28:41
  SPEAKER_01: Now this case you don't have an implicit understanding of the source system.
• 28:46
  SPEAKER_01: Here you need to gain an understanding of the source system and you would go there with like very traditional means you would look at tables and data structures and values in there but you would try to understand really is the business meanings the concepts the object out of that where exactly in which five tables does the material sit or does the product sit?
• 29:04
  SPEAKER_01: How is the product configuration actually manifested in that system?
• 29:08
  SPEAKER_01: How does it relate to POs, purchase orders?
• 29:10
  SPEAKER_01: How does it relate to invoices that type of understanding for many times very technical terminology that you have available in the system is there the predominant thing that you need to do to gain a semantic understanding of the system so you actually are able to move it into the target later on.
• 29:28
  SPEAKER_02: So you're saying that a lot of that stuff is kind of predetermined pre-defined pre-mapped but then there's still you know if you think of a pace layering approach right there are things that are very stable that move at a very slow clock speed which I think of ERP systems and there are things that move at a much fire faster pace you know and you have different layers of this and you know what happens in social media and marketing and uh messaging and sales and all that it happens very very very quickly right so there are distinct levers layers and and and levels of speed of change and then there are friction points in between right but where where you have a slower paced system interfacing with a faster paced system.
• 30:10
  SPEAKER_02: So at least a slower changing system right that's right slower changing system yes that's right that's exactly right so so I'm not completely clear on how that is considered or is it not something that's really considered has to be thought of from a different perspective.
• 30:26
  SPEAKER_02: So you're working on the foundation because part of what we've done in the past we've been brought in to build that semantic architecture and build that you know the controlled vocabularies and the classifications and the categories and all the different things and the data models to support ERP but also to support unstructured processes and you know knowledge processes because it really should be holistic.
• 30:46
  SPEAKER_02: You should have consistent organizing principles across all of your structured and unstructured information.
• 30:51
  SPEAKER_02: And we have a reference architecture that kind of talks about that, right?
• 30:55
  SPEAKER_02: But so so so how do you think about that and what's the practical advice for an organization that has to deal with this again uh uh ERP is much slower changing and it requires that stability it requires more of an enterprise um data architecture perspective it requires you know change management boards and decision making whereas there are other areas that are going to be much faster you know like much less oversight much less governance so how do you think of governance and decision making at these two ends of the spectrum the very slow moving the very fast moving is that a fair question I mean probably what what you not want but needs need to actually build is like you say you need to like build a ontology on top of what your business actually what your business actually is.
• 31:42
  SPEAKER_01: Yes and this is typically involving much more than just your ERP system.
• 31:47
  SPEAKER_01: I mean where ERP system has a certain gravity to it because ultimately this is where your production basically runs on this is probably also where your in one invoicing runs on is this where we where your CM processes run on most likely not these days right does all of your supply chain run there well no most likely not so I mean your whole enterprise architecture is of course much more complicated than that involves many many more systems and you do actually to make good decisions from a from a business perspective need to of course take into account all of these different aspects basically all these different systems and many times approach is being taken here is that you actually take the data out of relevant data out for decision making out of these systems of record put it into a data lake that is infledged with the ontology that you have as a company that's something that you build by yourself right that's your actions that's your objects that's your events that are basically happening and you build it as an abstraction on top of what your systems actually imply.
• 32:47
  SPEAKER_01: And most of the times it's a much more simpler view abstracted away from the technology from the technology that basically sits underneath and then what you need to do is of course extract the right data out of the systems and bring them into those data lakes to support the decision making processes and level out with this the different speeds that the different systems basically have.
• 33:08
  SPEAKER_02: So that is what the what the typical approach is is that recommendation no I think that's just a fact how how like enterprises act basically yeah yeah no that's those are those are valid points we like to think about and we just have been writing about this in our recent webinars on uh AI and and we have one coming up on governance next week we talked last week about information leverage points right where you're looking at everything is an information flow in the enterprise output is another person's input and at the end of the day you're taking in signals from the marketplace you're taking in information money is even information right we don't walk around with gold Kruger ends and hand them to people uh we take in ones and zeros and then we put out information even products even if you're a product company uh raw material plus information is a product right so a design I like to say my magical devices sand metal and oil all very cleverly arranged right my my my iPhone right and uh and and so that's you know hundreds of thousands of person years of knowledge and in and and information so when you think about what we're trying to do with all of these things are trying to speed up those information flows I call that the information metabolism of the organization.
• 34:18
  SPEAKER_02: And then we're looking for places where there's friction right and it's and sometimes when you're building a semantic architecture you're not necessarily solving one specific problem.
• 34:27
  SPEAKER_02: You're solving lots of little problems.
• 34:29
  SPEAKER_02: It's like you're getting pecked to death by ducks, right?
• 34:32
  SPEAKER_02: When you have all these information friction points where there's a manual process here or there's a translation here or this person has to wait for this data to be validated or you have to go in and check for compliance or you have a manual process or pro or a spreadsheet, whatever it might be.
• 34:48
  SPEAKER_02: But those are little things that in and of themselves don't look like much, but they add up to a lot of friction throughout the organization.
• 34:54
  SPEAKER_02: And there's kind of a fractal nature of processes within the organization and lots of information flows that have other information flows that have other information flows, right?
• 35:05
  SPEAKER_02: So when you build that semantic architecture and when you when you modernize these environments, you're you're speeding up all of those flows of information and you're letting people that way you can get to products to market faster, you can react to competitive threats you can react to changes in customer needs and all of those types of things.
• 35:23
  SPEAKER_02: And that's what we're trying to do with you know with normalization of data and and architecture as well as implementation of AI and AI initiatives.
• 35:32
  SPEAKER_02: So when you're thinking about you know this all becomes very complex because we have more interconnected systems, we're building more autonomous systems, you know, risk management becomes more challenging.
• 35:44
  SPEAKER_02: What have you seen work to kind of maintain control, transparency and accountability when you're starting to look at at these critical enterprise processes maybe that's a little bit outside of your focus or your baily weight but maybe you could speak to that from what you've seen when you're at these organizations and these are such critical processes and critical projects that are big investments.
• 36:06
  SPEAKER_01: Yeah I mean there's various ways of how to tackle that I mean some organizations just let it flow basically and have little governance over the thing right and over how processes evolve and how they basically lead to all the constructs that you just referred to right manual process in between everything basically everything is like a network right it's it's like basically a big flow with many decision points some manual some automatic some are structural some are governed some are basically not and some companies basically let it flow in specific areas and other companies try to basically gain try to stay on top of this right and then what what happens is there is this indexing or this redocumentation projects that are basically happening and that's like hey let's put all of our processes into one system and have them well documented and all all that's all that sorts of stuff um what I what I mean I mean my experience many times these are one-time efforts which are happening and then after they are there everybody is like tapping on their shoulders like yeah we did great and if they didn't great they well do like they did great basically and then from that point on what most of the time is lacking from my perspective is the governance downstream that really ensures that the consistency that you might have achieved or at least the improvements in consistency that you might have achieved they decay away pretty quickly once you lose sight of what you're actually doing.
• 37:24
  SPEAKER_01: So if you don't with a change also establish a governance if you didn't have it chances are you didn't right and you didn't establish it with a process chances are you did this for once and you will not have the benefits for long in your in your corporation.
• 37:40
  SPEAKER_01: So I would say rather than re-architecting the whole thing trying to stay on top of everything try to identify what the critical processes really are that if they are not exercised correctly really hurt your business and not everything is at the same level there right there are side process there are variants which are I'm not saying insignificant but maybe they are just like rare rarely existing and they are not meant to like completely automate in this particular case try to find where the value really lies or if you can't find where the value really lies try to find where your risk really lies.
• 38:15
  SPEAKER_01: Think about the last escalations that you basically had because of a system not working of a process being stuck of a decision that could not have been taken or that wasn't taken basically in the past identify those points find a remediation for that.
• 38:28
  SPEAKER_01: I'm not even saying like what the remediation actually is but establish a governance on top of that at the point that you identify so you're sure that in the future this thing doesn't break.
• 38:38
  SPEAKER_01: If I take an analogy from a software development perspective there's such a thing as test driven development and test driven development basically is the practice of you stand up the contract of how it should work before you actually do the implementation.
• 38:50
  SPEAKER_01: Now despite this sounds very very tempting to actually implement few organizations do that really yeah because it comes with some cost it comes with less spontaneity and you need to do a lot more thinking basically up front.
• 39:03
  SPEAKER_01: So many organizations do actually choose to do another path but if you find a critical error you want to make sure this thing is going to be tested in the future so at least that thing does not happen again.
• 39:13
  SPEAKER_01: That's the bare minimum of what you need to be doing.
• 39:16
  SPEAKER_01: So of course like everything in life it's about finding a balance basically it needs experience to find the balance uh at this point the only thing that I can really like solidly pledge for is like when you find such a thing make sure you establish a governance that makes sure this stuff does not fail anymore in the future and you have other problems to take care of next day.
• 39:36
  SPEAKER_01: Not the same problems again.
• 39:38
  SPEAKER_02: Yeah yeah and I think the the governance piece is so important because there are different tiers of risk right and you need to make the punishment fit the crime you need to make sure that you have the right level of oversight at each tier and building libraries of use cases to say what good looks like right so when this is done what this what's this going to do and that talks about the piece of functionality is that it can't fail here.
• 40:01
  SPEAKER_02: Well what's it supposed to do right what's the what's the positive of that the the affirmative of that to say this is what it needs to do and then we have to test right we have to have a baseline to say what's the current state now we have our intervention now we test again and that becomes part of our KPIs.
• 40:17
  SPEAKER_02: And so establishing those KPIs across the board at the different levels of granularity are really what what governance is about as well as having the right level of decision making so the right structures of committees working groups task forces right and then the right working agendas and and then the right measures to understand where people are doing that.
• 40:37
  SPEAKER_02: So governance I totally agree is a very important area that is our next webinar and just going to pitch that here although this will probably not be posted by uh the webinar so the webinar was on March 11th I'll speak in the past so people can look at that but where else can people find information about you and about your organization we'll put stuff in the show notes but how how do people uh get in touch with you?
• 41:00
  SPEAKER_01: Well of course naturally you can get in touch with with with LinkedIn right you can find stuff on the company's page what I would recommend though if you want to see like how we actually move data no even better how customers give statements about what their journeys were what the the challenges and how they actually overcame it overcome it with or with us out us as a company here or with our platform I can highly recommend you coming to Transformation World which is a key event it's happening once a year this time it's in July 8th and 9th if you happen to be in Germany or you can arrange it it's a great conference actually to go to and re life get experience from customers because that's what it's all about.
• 41:41
  SPEAKER_01: It's a community event right it's where customers speak about their journeys about their challenges and how they actually solve them independent of the Agnes uh technology Nazi and so that is is there a virtual version of that is or is it just face to face?
• 41:54
  SPEAKER_01: Is it just like well I mean it is an in-person version but of course we are providing all the contents the sessions and stuff like that also of course for past transformation worlds up in YouTube on our channels and so on so forth.
• 42:06
  SPEAKER_02: Any final words for executives in terms of what they need to look at what they need to think about what they should be prioritizing moving forward.
• 42:14
  SPEAKER_01: Yeah so generally speaking when an executive find a problem worth solving right don't find the easy problem right which is like like giving you like quick opening kicks and giving your organization a quick boost find something that really matters right in the end and try to either put an automation on top of that put a government on top of that and if that's a transformation project in the end that modernized your systems so be it.
• 42:37
  SPEAKER_01: Is it something else then be it something else basically but try to make a meaningful impact by looking out the organization holistically and find the spot where differentiation really makes an impact.
• 42:49
  SPEAKER_02: That sounds great.
• 42:50
  SPEAKER_02: Dominic thank you so much for your time today it was great to have you I appreciate your your thoughts and your insights thank you so much for having me.
• 42:58
  SPEAKER_02: And thank you to our audience and we will see you next time at the next early AI podcast.
• 43:04
  SPEAKER_00: Thanks everybody thank you for joining us on another deep dive into AI innovation.
• 43:10
  SPEAKER_00: Tune in next time when we introduce another industry expert and discuss how to maximize AI in your world.
• 43:16
  SPEAKER_00: The Early AI podcast is sponsored by Early Information Science Infector to learn more visit early dot com that's E A R L A E Y dot com Thank you for tuning in