In this episode of the SAPinsider Las Vegas 2025 podcast, host Robert Holland speaks with Dr. Berg of UNC Charlotte, a seasoned SAP expert and educator, about preparing the next generation of SAP professionals. With over 30 years of SAP experience, including pioneering work in SAP HANA and BW implementations, Dr. Berg now teaches applied machine learning, AI, and SAP systems at UNC Charlotte as part of the SAP University Alliance. The discussion explores how students are being trained not only in technical skills like Python and R, but also in critical business processes like order-to-cash and sales reporting. Dr. Berg emphasizes the importance of understanding data context, the evolving role of AI—including generative AI—and the benefits of building and training models directly within the SAP ecosystem. He also highlights the scalability and cost-efficiency of using SAP’s integrated AI development tools. Finally, he champions events like SAP Insider for staying current with innovations, learning from real-world implementations, and discovering new possibilities across the SAP landscape.
LV Podcast Dr. Berg_Trimmed
0:00
Hello, I'm Robert Holland, and this is the SAP Insider Las Vegas 2025 podcast.
0:05
Thank you for listening as we speak with SAP insiders and industry experts about their experience in the SAP space.
0:12
In this episode, I'm speaking with Doctor Berg of UNC Charlotte.
0:16
Doctor Berg, tell us a little bit about yourself and your role.
0:20
Yeah, it's been about 30 years putting in SAP system and HANA migration.
0:25
I wrote the first four books on introduction to SAP HANA and basically did a lot of my claim to famous or myths of claim.
0:33
I did the second B double implementation in the world ever at Ericsson in 1998.
0:38
I should have gone live a week before.
0:42
Wow.
0:42
So you have quite the history with SAP.
0:46
So you go back quite a ways, but what is it you're doing now at UNC Charlotte?
0:50
I'm a clinical professor, so not theory, but actually teaching people how to do do things.
0:55
So I teach Python, R, I teach machine learning and deep learning to both computer science students and business students.
1:03
And then I also teach at the SAP University Alliance.
1:05
So University of North Carolina is part of the SAP University Alliance, right?
1:11
Well, that's, that's so you're basically teaching the future of SAP in terms of the people that are going to be working on the systems.
1:20
People are going to be using the systems and you're bringing them up to speed so that they'll be able to effectively do their jobs in the future, right?
1:29
Yeah, that's the idea.
1:30
So we have some classes that are more in the University Alliance in the functional side, not in the College of Business.
1:36
And then we have people who do development, actually building neural networks and decision tree and random foresters of.
1:42
And we have that in the computer science department.
1:45
Fantastic.
1:46
So you're part of the SAP University Alliance, or at least the University of Carroll, University of Charlotte is, and you're teaching topics like SAP to newcomers.
1:58
What is it that they're most interested in learning about when it comes to SAP?
2:04
Well, initially they're all interested in building models, coding and C models and actually see their accuracy of the models.
2:11
And it's kind of like a baby when you go to machine learning.
2:14
You can see they're getting smarter and smarter as they keep running through what they call epochs and but then they realise the hard part.
2:21
Do you understand the data?
2:22
You know what the data field mean?
2:24
And sometimes you can build something we call A championship of the bleeding obvious because these everybody knows that these two fields are related and they think it's so cool.
2:32
It's simply because they don't know the data.
2:34
So if you want to work in this field, you need to learn about process flows.
2:38
You need to learn about like order to cash.
2:40
But they said the student come in and says I need to do sales reporting and we turn around and says for who just well, sales reporting just you want to sold, to ship, to build to a payer, Would you like to have the sales order or what we delivered, what we invoiced, what's an accounts receivable or do you want to go to general Ledger and see what we actually got money for?
2:58
And they go like, what does not compete?
3:02
So then we have to come take a giant step backwards and understand business processes, document flow and realize that sales data offset reporting comes from so many areas.
3:10
So, so many aspects.
3:12
Yeah, It's when you start thinking about the SAP data ecosystem, There's so much data across so many applications and it's, yeah, there's no what there's no, there's no way you can say I just want to do sales reporting, for example, because, yeah, it comes from everywhere.
3:34
I mean, there's, there's, it's funny, when you were first implementing SAP HANA systems, you could probably run, you know, an ERP system on about 200 megabytes of data.
3:44
And now it's substantially more terabytes is many people have TB system used to be a TB club and still remember them back in the 2000.
3:54
It was a bragging right.
3:55
You had a TB of data.
3:57
Those 10 companies in the world.
3:58
It was so amazing.
4:00
And today it's like small startup companies have a TB.
4:03
I mean, I was with, I went to a, I went to a session with IBM and they were talking about how they had converted Pfizer's SAP is for HANA implementation.
4:13
It was the largest single instance of SAP is for HANA converted.
4:18
And it was like a 200 terabyte model.
4:22
Now they cut it down, moving to SAP is for HANA.
4:24
I think it's like 40 terabytes now, you know, but it's like, yeah, the data volumes are just incredible compared to what they used to be.
4:31
And that's before you talk about companies like Walmart's who operate into the petabytes.
4:35
Yes.
4:36
Wow, the amount of data that they track or the Googles or the the Amazons or the Microsoft, a lot of the consumer package companies you find find they have a lot of point of sales data and they keep quite a bit of it because they want to do cycle of trend analysis and see business cycles in 6-8 years and compare them.
4:54
So you wind up with really cool capabilities, but now you've got to make sense of it.
4:58
Then get somebody into actually build stuff as well.
5:00
Yeah, absolutely.
5:02
So you're teaching students as part of the SAP University Alliance.
5:09
Where is it they're going after?
5:11
They're coming to your classes and they're learning about topics like, for example, what it really means to build a sales report.
5:19
Well, if you're only in the functional side of IT, a lot of them wind up in business.
5:23
So they wind up working with accounting majors and eco majors and finance majors.
5:29
But if you're more on the computer science side of it or the business information system people, then they actually wind up and developing stuff.
5:36
So they actually go up and fingers on the keyboard and they say, I now want to be productive.
5:40
I don't want to talk about it.
5:41
And we joke and says I don't want to be a Power Ranger living in PowerPoint.
5:45
I want to have, I want to code in Python And R and I want to do something meaningful.
5:50
And right now that's pretty nice.
5:51
My youngest son just finished his masters in applied data science and I was amazed the money he made serial experience straight out of school.
6:00
That was like science fiction money for me when I graduated.
6:03
It's a little bit different now, but it depends on the role and the job and that sort of thing.
6:09
But you're teaching, you're teaching students, You know, one of the things you said that you have, you know, you focus on our Python AI.
6:18
I mean, AI is so much now a part of what SAP is doing.
6:23
What, what AI lessons do you sort of see the students learning when you know they're and I think your session, your session here is about literally physically building an AI neural network, neural network.
6:40
I did, I did it live in the training in the session, show them how to build a neural network.
6:43
And then we saw that our baby trained and get smarter to predict machine failures.
6:48
We had like 470,000 machine hour data and we see, can we predict if the machine is going to fail the next hour?
6:55
And we did it 97% accuracy after we trained the model.
6:59
So it's not as complicated some people think it is.
7:02
When they get to see it, instead of just talking about it, they go like, wow, this is not that hard.
7:07
And I tell them this is what a junior level undergrad developed in class.
7:11
So when you get them, don't treat them like they don't know anything.
7:14
They may not know SAP and they may not know the business, but coding stuff like this, we take them head finger power skills straight out of university.
7:22
Yeah.
7:23
Yeah.
7:24
So there's, you know, they, they may not be SAP experts.
7:27
Yeah, that although they've had exposure to SAP systems, but they're using technologies that a lot of these organizations now want to incorporate into what they're doing as part of their, you know, business plans and business processes.
7:42
And we see the three things they need.
7:44
They need to be able to actually build something.
7:47
Then they need to have domain expertise.
7:48
So for instance, AI today is so big.
7:51
They have machine learning.
7:52
You have, you can go in for instance, like natural language processing, text analytics, domain sentiment analysis, optical character recognitions, computer missions, identity management, facial recognition.
8:04
You can go in to do predictive analytics.
8:07
You can go into prescriptive.
8:08
What should my production schedule look like?
8:11
So right now we tell the students we're going to give you the base foundation.
8:15
But at one point in your career, you need to get domain expertise and you're not going to be an AI developer, you're going to be an AI developer focusing on something certain techniques.
8:25
So how much I mean, you talked about a lot of different AI capabilities there.
8:31
How where do you sort of see generative AI coming into the picture?
8:35
That's an even more fun one.
8:37
So if if we went leaps about look just at chat CBT version 4, a lot of the hallucinations be golden in version one and two went away.
8:45
If you look at Gemini from Insta, from Google, they stopped, they put this filter bias filter, created some very embarrassing pictures that took it offline I think in March, April last year and then relaunched it.
8:56
And they they're not like incremental improvements, they're leaps and bounds improvements and that is very fun to see and, and seeing that SAP actually embraced it, sitting on top of the data, built a specific hub inside the BPT for generative data.
9:10
That's kind of it's not it's not just a me too, it's trying to be a thought leader.
9:15
Yeah, yeah.
9:16
And SAP is investing significantly, you know, in AII mean obviously there's business AI that they're embedding within their applications, but also there's the ability to connect to AI in, you know, so many different models, so many different ways through AI foundation, AI core in B to B.
9:35
So that's one thing I do like how they did, for instance, the Python interface.
9:40
When you develop, develop an SAP like open AI, you don't have to take the data out of the platform.
9:46
A lot of the stuff we wind up doing for traditional AI development, we suck up the data, we get giant data files, we need to stage them, then merge them together and ETL.
9:55
Then we try to download into the application development environment and it blows it up.
9:59
I got some souped up servers just to be able to do this right.
10:03
In the SAP environment, the data doesn't have to leave the HANA platform.
10:07
I can have the data there, build my model and train it the data and deploy the data model or the AI model back into the SAP platform.
10:14
It's the HANA that gives me the speed and more importantly also give me logging most systems out there, like if I'm in a bank and in order to be system a record hunt Sarban Oxley and must be able to log what the system does.
10:27
They're known in natural logs that exist in AI now that's what SP has brought the table and open AI as well.
10:34
Yeah, yeah, yeah, absolutely.
10:36
So what would you sort of set, what is a piece of advice then that you would sort of offer?
10:40
Because you, you talked about, you know, the fact that a lot of your students are, you know, doing custom development.
10:48
You've done a lot of custom development in the past.
10:51
So what's a piece of advice that you would offer regarding integrating that custom development while consuming SAP data?
10:59
One thing I do like is the, the ability to take the platform and open it up and spin up environments and shutting them down on demand.
11:06
For instance, when I build a model, my students do we always have to provision for the peak load.
11:11
So I build a model, I start training it and then just explodes in RAM and CPU graphical processing units.
11:18
But as soon as I've finished doing it, I don't need that power anymore.
11:21
I might need a fraction, 5% of it.
11:23
But now I have built a system that's 20 times more than I need just to train the model.
11:28
The cool thing with the open eyes, I can do that dynamically, it spikes up or when I don't need it, it goes down and I don't have to pay for it.
11:37
So that's kind of what we tell the student is don't do these massive AI data centers.
11:43
Some people have to do that, but in your companies, you don't have to do that because there's an open AI platform for you within the ESP environment.
11:50
Yeah, that's an important piece of advice because I don't know that everyone necessarily recognizes or utilizes that.
11:56
No, it's like I just did a session literally an hour ago and I had a room full of expert people, a lot of AI developers, and I asked them how many people know, are aware of this.
12:06
And I saw that 5-6 hands, how many people have heard about them?
12:09
I got 7-8 hands and I asked them how many people have deployed it like A1.
12:13
So we're still, we still have to get the message out that SAP is not the proprietor download platforms that you can't do all your code works, all your libraries, all your Python packages, everything runs inside it.
12:27
It's a graduate.
12:28
Do not have to relearn some new tools.
12:30
You just have to learn the data.
12:32
Yeah, absolutely.
12:35
So there's lots even for experienced AI developers to learn.
12:40
And it sounds like because you've been coming to this these sort of events, you know, as you told me from something like 2003.
12:48
Yeah.
12:48
So what would you say the benefits you would receive from coming to an event like SAP Insider, apart from getting to meet experts like yourself and learn new things about AI and and what would what do you think would make it even more valuable?
13:05
But first of all, the events have gone tremendously.
13:07
Used to be 70 or a few 100 people at this event.
13:10
There are thousands of people, there are hundreds of companies.
13:13
And so it's really fun to go back to the well and hearing other people because the environment and ecosystem, I got them so big.
13:20
So just go back and hear from other people, pick up and learn from them.
13:24
See what people really are doing, not just what's available, but a lot of people implementing.
13:29
It's kind of, I think it's a must do.
13:31
If you skip events like this and put your head in the sand and you kind of think, read some PDFSI think you're going to lose the boat completely.
13:39
Yeah, It really gives you the opportunity to hear from other SAP users about what they're doing, what they're experiencing, and then learn from that yourself.
13:49
And some of these companies that I just walked around the floor and took a look at the Conference Center here.
13:54
If the art of the possible, there really things that never even occurred to me that you should be doing.
13:59
And they have huge value proposition.
14:00
Like, there's one in here showing me a print management.
14:04
I never even occurred to me with that being issued.
14:05
But if you have statements and PDFs and stuff that needs to be mailed out, it's a huge problem.
14:11
That's really fun to see both petite firms and the big, big players as well.
14:14
Yeah, absolutely.
14:16
Well, Doctor Bert, thank you so much for talking with us today on the podcast.
14:21
I really appreciate you and your time.
14:23
You're welcome.
14:24
Thank you.