AI-Native ERP: Turning SAP from System of Record to System of Reasoning | The BizTech Pulse Podcast

**********
TRANSCRIPT
**********

Krishna Hari (00:05)

Hi, good morning. Welcome to The BizTech Pulse Podcast. On today’s episode, we are excited to welcome Vineet Moroney a trusted leader, advisor, a strategist, and a transformational expert. Vineet helps a enterprise and their team navigate the complexity of SAP ERP projects with clarity and measurable impact. Currently serving as advisor, he leads AI initiatives, ⁓ crafting scalable solutions that integrate AI into core business functions. His passion lies in leveraging AI, GenAI, and data-driven strategy to deliver real business value. Vineet brings a rare combination of ERP expertise, strategic insight, and forward-looking vision to the conversation. And today we will explore how organizations, especially mid-level enterprise, can unlock the power of SAP and AI to drive transformation and growth. Welcome to the podcast, Vineet. Just wanted to have your initial intro, if you can talk about yourself.

Vineet (01:11)

Thank you, Hari. Thank you for having me here. My name is Vineet Moroney. I was responsible for enterprise apps practice at a Big Conglomerate. But now I am advising customers on implementing their AI strategies, how AI can be leveraged in the processes and transforming their organization. That’s my core job as of now. And having a background in ERPs, I always look at how we can make ERPs more and more, I would say intelligent and more and more thinking systems rather than static systems.

Krishna Hari (01:46)

Awesome, awesome, awesome. That’s in tune with the trend currently. Let’s start with my first question. think you have spanned enterprise apps and now deep AI work. What inflection point convinced you that the next decade ERP would be AI-led? Looking back at waves like IoT, RPA, and blockchain, what made them stand out in the enterprise compared to AI’s current momentum?

Vineet (02:13)

So Hari, I would say that the real turning point was for me was not a lab demo, it was a customer pain point. One of my leading French customer invited 14 partners to pitch innovation because post-COVID their supply-chain was broken. And the biggest frustration was they could not promise accurate delivery dates. The ERP had all the data, orders, supplies, capacities. but it was stuck in fixed static rules. It could not execute based on the patterns which they had set. It was just executing based on the rules and there I found that there is opportunity. Why are the enterprise systems just execution engine when they could be thinking systems? And That’s why I thought that this could be a great opportunity to embed intelligence in the execution system. And that was the inflection point for me.

Krishna Hari (03:05)

⁓ That’s a great pivot, and I understand. I think as far as the real-time reasoning of the ERP is concerned, definitely, think That’s the way to go. That leads us to the next one. You said ERPs are still largely rule-based systems, What are the biggest business gaps this creates? In practical terms, what does it take to make ERP pattern-⁓ aware?

Vineet (03:30)

So the first thing is that ERPs like SAP sit on an enormous amount of structured data. But the gap is – the data is not being used to derive patterns or to make business processes predictive. The system is still locked into static predefined rules. That’s the biggest gap I see. What we really need is that to analyze this structured data with strong AI models, extract inferences and then feed those into the system as dynamic rules. That’s how ERP should move from being passive recorder systems to intelligent systems and drive agility for the business. In simple terms, it should analyze the data, create inferences, apply those dynamic rules and make sure that the business is More agile, more predictive.

Krishna Hari (04:23)

Today, that structure is not there in the core ERP to do that. That’s where coming to the next one, if a company has a legacy SAP, what’s your three-phase path to AI-native SAP? Where should leaders expect earlier wins?

Vineet (04:41)

So the three phases in my opinion is: First is the data foundation. Data in ERP needs to be harmonized whether it is master data or it is a customer data and so on and so forth. The second is to connect to external signals. Right now, ERPs are not connecting to the external signals which are there. So first building a data foundation, second after building that data foundation embedding the intelligence on that. So applying the model understanding the patterns, understanding the behavior and after that creating an autonomous loop what I called as which we called as a agent. So basically looking at this pattern then look at your data and where are the adjustment you need to do and the early wins could be let’s cash application. You can predict your cash applications, you can predict our project cash flows, our cash flows very easily. You know the pattern that this type of customer don’t pay on time. This type of project you don’t get ⁓ payment on time. So those patterns are there in systems. We need to sense those patterns, apply the intelligence on that and embed that intelligence in our process. That’s how I think we can make legacy SAP to an AI-native baseline.

Krishna Hari (05:54)

It’s a very good point. think if you see a few startups on the FinTech area are doing exactly the same thing. They are pulling the ERP data, running their AI ML intelligence on that, and also getting external data like credit scores or the business news which comes in, basically then process this information and give it

Vineet (06:14)

See you.

Krishna Hari (06:20)

back to the user, to the same ERP user saying that, this is what is happening with this user. So the probability of payment is maybe 50%. Earlier you thought it was 80%, now it is 50%. So the risk portfolio comes into the picture. Coming back, I think with the AI wave which is happening, the other important topic is on the software development area, enterprise software development. “Vibe Coding”, I mean, I know ⁓ you had written a very nice white paper on that. A lot of good information, good research has gone into this. I wanted to talk about it. You’re arguing coding isn’t a job, code architecture is. What becomes the core skill set? What is vibe coding? If agile alone won’t cut it, what replaces it?

Vineet (07:04)

Absolutely. So, vibe coding is nothing but you need to set the intent, give the context and AI will generate the code for you and orchestrate for you. The danger is invisible tech debt. If the architecture is ignored or guardrails are not being used or reuse function is not happening, so it can generate debt So, in the AI-assisted development, code quality equals to prompt quality. If your prompts are inconsistent, then your code quality will be inconsistent and that will not be as per the architecture guidelines and so on and so forth. So, that makes the role of a code architect pivotal because role is designed to architect, reuse and enforce governance on the prompt-based engineering, AI based assisted tool. So, in short that if we do not have this role, probably we will create a system which will have lot of technical debt with lesser reusability and very difficult to debug because now the code is generated by AI. So from the delivery perspective, agile still will work, but the processes now need to be moved towards AI-augmented delivery. So basically, if you look at it there are no new roles which I am talking about – a solution architect, code architect. Similarly, you need to have AI + human review gateways in some of those things, to make sure that the things are evolved properly. So basically, you need to adjust your process to new methodology. And That’s what I believe that if we have to deliver successful vibe coding, it’s all about orchestrating all these things together, the role, the prompts and the quality of the prompts that will make your code or your product much sustainable than what you are developing just with prompting without thinking.

Krishna Hari (08:57)

I’m thinking, does it have an impact on the development time of a particular project than what it was earlier and what it is going to be now because of this?

Vineet (09:07)

Absolutely. There is research which says that the development time can be 30 to 40 percent reduced. But if you look at the tools which are there today are code-biased tools. Now whichever tool you launch it is code-biased tool. SDLC is not coding, coding is just 25 to 30 percent. Right now the first thing comes the requirement, whether the requirement is consistent or not. That requirement needs to be groomed properly so that the features are properly developed or not. So in my opinion, AI should be applied to the entire SDLC lifecycle, not only coding. Whereas I see that most of the time we talk about coding, which is 30% job, and 70% job we are not talking about what sort of thing we need to do. So I believe the right requirement will generate right code because there are right prompts being developed. So That’s where I think the focus should be the entire SDLC life cycle and if you look at the entire SDLC life cycle, there is a 30% to 40% reduction in cost, as well as now research says that 33% to 44% reduction in the time to market.

Krishna Hari (10:15)

That’s amazing. That’s amazing. And also, I think that comes to the next question. Where can 90% to 95% automation realistically happen today? What must stay human in the loop? How do you design controls in the AI area?

Vineet (10:33)

So, human in the loop is going to be an important gateway because AI + human will deliver miracles. If you just trust on AI it will not. So, human reviews are important. So, every where if you look at the SDLC life cycle also, if the AI can do requirement grooming, somebody needs to look at saying that requirement is now properly defined and say that yes, the human puts in approval for that. In the SDLC life cycle, somebody creates the technical document, the technical document needs to be reviewed by another human or another AI agent. Then when the prompt is created by the tool or the code is created by the tool, there should be a peer review which is done by either by an agent or a human in the loop. So unless there is a human in the loop, there is a possibility that we will deliver a code which may not be in sync with the architecture which we want to guide through and it will not be in line with the technical documentation which I want to create and I want to align the code to that technical architecture. So That’s where I believe that human in the loop is going to be important. I call it as a digital twin. There is an agent and there is a human. So we are not eliminating human’s role, but we are enhancing human’s role by providing the right agents.

Krishna Hari (11:49)

That makes a lot of sense. Coming back, I think, this is again, I know there’s a news in the industry that all the technology companies, especially Microsoft, (Facebook) Meta, they’re all, I mean, laying a lot of developers off. Either they are over-hired or they feel they are the first… taking advantage of this AI platform in development and stuff like that. Maybe they are doing that pretty faster than any other industry. So there is a fear among even all the developers or the computer science graduates now, whether they will get a job or not. So can you address that?

Vineet (12:28)

Absolutely, I think if you look at the trend, the person or the job is getting eliminated, but that job is getting recreated with a human + AI. So, the person who doesn’t know AI will get eliminated, but the job gets created, which is human + AI. When I talked about coders are getting eliminated, code architects are the roles are getting created in the system.

Krishna Hari (12:49)

getting creative.

Vineet (12:51)

So for the new software engineers, I think what they need to do is that they need to is adopt this new way, make sure that they are fluent in the AI tools being used. You don’t have to innovate, but you have to use these tools very effectively in your operations and the day to day work. And That’s what will make their job much simpler and much better than the person who doesn’t use AI.

Krishna Hari (13:15)

Absolutely. Absolutely. I’m with you there. The next question is, this is towards SAP Joule and BTP and the model layer. How should customer think about SAP Joule + BTP when most frontier models live outside the ERP?

Vineet (13:32)

No, this is a good question. I am, first of all, big believer of “invisible AI, visible results”. What I say is that AI is often mistaken as a chatbot because the avatar of AI was a chatbot. So everybody said that if I have to showcase I have got AI capability, I need to create a chatbot or a copilot and Joule is a copilot, right? And Joule is a copilot for developers, consultants and business. It’s a good start. but it’s not embedding into the core processes. So if you create that, the co-pilot which will help developers, maybe consultants but for the business users or businesses the intelligence needs to be embedded into the processes and That’s where I think everybody needs to look at that. So SAP also use these all those models which are developed by OpenAI and others. So it’s not that they are not using it, they are using it. The only thing is that Joule isn’t still an open system, right? When I called as open system connectivity – there is an agent-to-agent collaboration which can happen, which is not there in Joule. I think over a period, think SAP is now getting there. They will have this capability that agent-to-agent collaboration will happen. A deeper intelligence into supply-chain or procurement is still awaited. So I see potential in SAP because SAP has got such a humongous data and insight. SAP will move over a period from a co-pilot to agentic supply-chain or agentic procurement and moving from assistant to autonomous solutions and That’s what I believe SAP will do over a period. So framework would be Joule would be AI-native co-pilot, BTP will be the extensibility layer and the frontier models or open source model they will consume to create the solutions. which will be AI-native solutions – that is what I believe as of now.

Krishna Hari (15:24)

No, no, I agree. I agree there because I think I have seen a few on the agentic AI side, a few smaller firms, startups who are focusing, as you said, on the supply-chain, on the financial transformation. And they go much deeper. They go much deeper and they are more aligned with the process. That’s where I think I feel maybe SAP will do couple of acquisitions. I mean, that way I think they can speed up the whole process. Coming to the next question, this is more for a mid-market manufacturers, 300 million to 2 billion. With a lean team, what open source stack and cloud choices let them prototype faster without overspending? What’s your 90-day pilot blueprint look like?

Vineet (16:14)

Okay, so ⁓ mid-market firms – first of all – they can’t invest in massive models and not also in big data science teams right now because if they invest in massive model and big data science team the use case or the ROI cannot be even met. So the best case strategy would be that don’t try to build everything – look at the innovations which are happening in the ecosystem. Partner with the innovators rather than doing everything yourself. So if your strength is let’s say your strength is in particular area, look at that area, look at the partner and innovation happening in that area, partner with that, and then use those models or those innovations on your data. So basically you bring AI on your data rather than building yourself AI from the scratch. The focus should be that

Krishna Hari (17:04)

That’s right.

Vineet (17:05)

those 20% problem which will have 80% business impact. If these 20% problem needs to be addressed, we should address those 20% problem. We should not address the peripheral issues because “AI is good to have, I have got AI” type of thing we should not have. So, if somebody wants to develop on their own because now I did that earlier then there are lightweight LLMs like Mistral. Now, Mistral is saying that they will make it free for everybody then there are vector databases and then orchestration layer like LangChain or n8n which can be used. From the 90-day pilot I would say that – pick the use case which is going to be impactful, do not pick something which is not going to be impactful. Prepare on small set of data set, fine-tune with the partner, measure ROI. If you look at this strategy, you will able to get the ROI and you will get able to get the adoption of AI in your organization. So, don’t look at the technical nuances – which model, which database and so and so forth. Look at the business problem. Look at what is the innovation happening there. Partner with the innovator, use your data, fine-tune the things and then create a prototype. and it should be like POC-to-pilot in 90 days – we should able to get the result in 90 days.

Krishna Hari (18:25)

Absolutely. Absolutely. It’s a very good model to follow, especially for the mid-level customers. Coming to the next, can you share a playbook or a before-and-after story, given your experience?

Vineet (18:40)

So let me go back when I, where I started, right. One of my customers came back to us after COVID – their supply-chain is broken. And because their supply-chain was broken, they were not able to predict the delivery date to their customer. And that created lot of, restlessness in their customer – every time they were changing the date. They came to the 14 partners and then we said that, no, we can look at the problem and we can try to find out in two ways. So first thing is back to basics. So SAP has got so many rules. We looked at all the rules and said whether these rules can be made a little dynamic. That’s where we looked at the, for example, inventory buffer stock, looked at the reorder point, looked at the lead time of our particular vendor. and looking at the pattern we could say that no these are the thing which needs to be critically addressed in the system. So That’s where we said looked at the data and said that no let’s clean up the master data using the latest pattern which we are evolving from the data. But That’s just optimizing your current state of supply-chain – what is important is not optimizing but forecasting in future -if you want to rebalance your supply-chain what insights you have got? So, then we looked at AI. So one is fixing the basics and then we looked at the AI which will bring the insights and say that if you want to rebalance your supply-chain you need to transition from this supplier to this supplier for these items you need to do a vendor-development activity for these items. So, basically My playbook is simple. First is focus on back to basics, embed AI inside your process, not just on the top. And that is what helped the customer I think out of the 14 partners, they awarded us the contract and we helped them in improving the predictability date. Earlier the ATP (Available-to-Promise) adherence was about 80% which we took to about 92% by implementing insights.

Krishna Hari (20:43)

That’s amazing, amazing. Great story. I think That’s definitely… Is there something you can create a model out of this? For example, on the supply-chain, I think whatever additional data you captured, historical data, I’m just thinking here how to create that into a small product for others in the marketplace as well.

Vineet (21:05)

Absolutely, I think this is doable, I think there is lot of data sitting in SAP. The only thing is that we need to look at the pattern and we need to relate those pattern and we need to know how the SAP works – what sort of ways we need to tune the – like no lead time for a vendor or let’s say reorder point and so on and so forth. So we can look at this and we can create a dynamic model which we can run every weekly and fine-tune these parameters based on the findings and make sure that we can make the process, supply-chain process more agile, more predictive and making sure that no we do not overstock inventory but you adjust the process based on the patterns. Now the third thing is getting outside signals is also possible, with Gen-AI

Krishna Hari (21:57)

Mm-hmm.

Vineet (21:57)

so many market fits are there, you can get those outside signals and then embed that outside signals into that and create, I would say that create a solution which will embed the outside signal, use the internal data and create a harmonious solution. By the way, we developed this and we won the first hackathon of SAP using this solution. So That’s what I said that it is… it is on SAP’s

Krishna Hari (22:22)

Wow.

Vineet (22:25)

market site where you can look at this solution.

Krishna Hari (22:28)

Oh That’s nice. That’s nice. The next thing is, I know you did mention that you had an analyzer tool set created earlier. If it is for the AI world, what will be your version to do if you can share your product roadmap or whatever it is?

Vineet (22:44)

Absolutely. So, I remember that I was in a customer meeting and in Frankfurt and then the idea struck to me that it is time to create a tool which will analyze the functionality, analyze the data and create a real, I would say processes which are going to be there for a customer. So, That’s where I created a tool. which was earlier name something called as a FUNDA which is – functional and data analyzer. Over a period it has gone through multiple iterations and then we continue to improve on that. But the tool which was developed earlier was again a rule based tool. We used to look at the configuration customization in the SAP, from that we used to lay out the process and then we used to look at the data. the transaction data and then we used to correlate that and say “This is what is the process and this is what the data says” – and also we used to look at the what we called as a customization SAP like user exits and so forth and try to relate that and try to find the vulnerability index. So, all these were based on the knowledge of the people of the system. Today if I have to do that I do think I will do in the same way. We have got now so much of data using that data you can find the patterns, you can find the processes, you can find the I would say bottlenecks in the system and then advise customer saying that this process needs to be reimagined like this or this process needs to be reinvented like this and so and so forth. So, I will use the data which is there, will analyze the pattern – looking at the pattern we will to find out the vulnerabilities where the things will fail – and that model can be self-learning over a period and make sure that now we can give insights to the customer that what is happening in their system.

Krishna Hari (24:37)

Amazing. Amazing. So that will be the next AI tool which will help ⁓ the customers. Definitely we should work on it. The next question is – One of the massive change which is coming because of the AI. What should leaders do in the next 6 to 12 months? What is your thoughts on that?

Vineet (24:43)

have to. Yeah. So, when it comes to AI, if you look at it, the first thing is that it is not the technology which is not working, it’s the people behind that. So, first and foremost thing which comes to my mind is building AI-first culture. As I said that strategy is useless if culture does not exist. And the tougher resistance often seen is in middle layer. They face customers, but they resist change. So, win them over by enablement and recognition. That is what I would say that build this AI-first culture. The second thing is that for some of those roles which you have developed earlier – redefine those roles so that there is a career path for every individual. So, Developer becomes Code Architect, SAP Lead becomes AI Product Owner. FP&A Leads becomes AI-assisted Planner and so and so forth. Third thing is the rebuild process, no short on the cycles, AI + human get reviews, KPIs on exceptions. These will help actually to build what he called as the AI adoption in the organization and it depends on the people and if there is no AI-first culture everything will fail.

Krishna Hari (26:05)

Well, very well said, Vineet. I definitely agree on the culture side and also the renaming or renaming of titles in accordance with the AI way definitely will make a huge change. Coming to the next one, in terms of the next 5 years or 10 years – especially 2030, where is our ERP headed and what shaped your philosophy?

Vineet (26:29)

So, I believe by 2030 if ERP does not embed AI, those ERP will be laggards. So, vendors or the customers will look at that systems which are thinking, which are learning rather than system of records, right? System of records is there, but this is what people will think. So, my opinion is that in 2030, we should not lay AI on top of the current things and say that now we are AI-native ERP. It needs to be embedded into the process itself. So operations where human is in the loop only in the edge case, open connectivity, agent-to-agent collaboration happening. That is the way I see that ERP will able to add value to the customer. And if personally I think that now

Krishna Hari (27:03)

Thank

Vineet (27:23)

if the OEMs are not inventing, the customer will reinvent using third party tools. So, my philosophy is simple, right now, that AI is here to actually enhance the usability of the ERP moving the systems from system of records to system of reasoning to system of thinking to system of learning – and that will make transformation and better adoption. See what is today’s problem in ERP is also that adoption. Whenever you see most of the time everybody will say that no ERP is not adopting the organization. It is okay. I fill in data. I get some output, but it is not helping me to do my business. With this change which is no system of records, so system of thinking to system of learning, the adoption will be much higher.

Krishna Hari (28:11)

Actually, I wanted to bring one particular analogy with ⁓ applications like Uber and real-time applications, right? Even Airbnb, Uber – where you have a real-time, for example, if somebody calls a cab Uber or Lyft, a lot of real-time. ⁓ data is churned and given back to the customer in a fraction of seconds. They make the matchmaking, everything is real-time. I feel ERP should be like that. Every ERP should be like that in terms of… ⁓

Vineet (28:36)

Absolutely. Yeah. No, absolutely. I think that ERPs need to evolve themselves. I would say that, if they can’t become Uber like real-time sensing, but they can definitely become thinking systems, right? They can definitely give the signals where are the things going wrong, where are the issues happening and they will able to, even if that is being done, probably we fine-tune our rules in the system. once in a week or once in a month – that is okay for the enterprise grade systems.

Krishna Hari (29:22)

Absolutely. So coming to the, mean, yeah, we just wanted to go into the rapid fire round. I just wanted to, it’s a short answer. I expect from you. What one book or podcast you always recommend to colleagues or clients.

Vineet (29:34)

So book – Execution. Because execution is a key. You can have vision, you can have a strategy, but if you don’t have execution, nothing will happen.

Krishna Hari (29:37)

execution. Absolutely. Any podcast which you listen regularly?

Vineet (29:47)

So, I listened to Kamath’s podcast which is, I feel that it is great insight brings to the human life and so and so forth. 

Krishna Hari (29:57)

Do you prefer early morning or late nights for your best work?

Vineet (30:01)

I avoid early morning because those are my exercise hours. So I believe that now we need to exercise for one hour and therefore late night is okay for me.

Krishna Hari (30:09)

Okay, okay. If not technology and AI, what field would you have pursued?

Vineet (30:14)

I would have pursued singing. I am not a great singer, but I love singing.

Krishna Hari (30:19)

I didn’t know that. Nice. Karaoke sessions and stuff like that. Nice, nice, nice. And what’s one daily routine or habit that helps you stay effective.

Vineet (30:23)

Absolutely, I love Karaoke positions. So, one is learning. I learn continuously. That is where I feel that when I tell my son, daughter – do not leave learning. So, learning keeps me updated on my skills – what is happening. And I would say that now unless you learn every day, I try to find that what I learnt today at the end of the day. One thing which I learnt is what I call it is I had a successful day… or I had pathetic day. So successful day is one learning, pathetic day is no learning.

Krishna Hari (30:59)

That’s a great one. I actually, I don’t know that you read this book called Atomic Habits by James Clear. He basically says, I mean, every day if you do something different or learn one, even a small piece, I think that adds up cumulatively. And so what’s your go-to activity to relax outside of work?Vineet (31:15)

Go to activity to relax is watching good movies. That’s one thing I do right now. I like watching good even listening some of those podcasts, which are not technical podcasts because that’s relaxing. Yeah. So that’s what I do.

Krishna Hari (31:33)

okay. That’s great. That’s fantastic. I mean, ⁓ thanks for playing along, Vineet It’s great to see the personal side ⁓ behind the professional capabilities of yours. basically, this is actually wanted to wrap this podcast. If you are a bit size firm looking to scale ERP and with AI, here are the 4 starting points. One is – Automate high impact, low complexity process. 2. Leverage AI driven insights from ERP data 3. Integrate AI into your user experience. And the 4th one is, yeah – Feel free to fix a ⁓ 30-minute free call with our ERP advisory experts. We’ll be more than happy to help. Coming back, I think to our listeners, if you enjoyed today’s podcast, don’t forget to subscribe, rate and comment at The BizTech Pulse Podcast in your favorite platform. Every bit of feedback helps us bring you more conversation like this. You can also follow us on LinkedIn, Spotify, and Apple platform for updates. If your conversation with Vineet sparked new ideas or reflection, we would love to hear from you. Please drop a comment on the comment section. Thank you again for tuning in. Stay curious, stay innovative, and we will see you next time at BizTech Pulse. Thank you

Vineet (32:54)

Thank you.