AI Leadership Insights: SAP’s Advancements in Agentic AI with Manoj Swaminathan

ai video series. Today we're with Mano Swami, nothing, who is, um, general manager and Chief Product Officer for SAP Business Suite, the finance and spend portion of that at least. And we're gonna be talking about, well, what is going on at SAP Sapphire this week?

'cause well, there are agents everywhere. Manoj, welcome to the show. Thank you so much, Mike, for the opportunity and, uh, really nice talking to you as well.

So, SAP has been, you know, driving much of this conversation around AI agents for a while, but this week it seems like you're building out an ecosystem, includes the foundation and a and a lot of different agents that are already maybe, uh, pre-trained to do certain tasks. Kind of walk us through this at a high level. Absolutely, Mike.

So I think, we'll, we're launching four key things and, uh, you know, what you've heard from us, um, today, I'll quickly touch upon each one of these things. So the first and foremost is, you know, we've completely reimagined our enterprise applications with omni person business ai, right? And this is intended to actually give every enterprise up to 30% gain in terms of their productivity.

So what does omni person business AI really, really mean? Uh, right? So it's it's omni person jewel that puts the power of generative AI in every user's hands.

The operating system for AI development is something that we are launching as well as what we call as an AI foundation that transforms the way SSO every engineer within an enterprise would build the deploy and scale the AI solutions. And we're also launching a network of autonomous decision making agents. So that can actually work with the real-time business data.

And it's also orchestrated by JUUL that turns agentic AI into sort of tangible outcomes. We have these agents across our suite starting from the financials through supply chain, human capital management, as well as our commerce suite of applications as well. And in terms of is how they are prevalently available supporting the automation of, uh, you know, the set of tasks that water typical persona could possibly be doing.

I think people understand at least the core concept of an AI agent at this point, but people are kind of scratching their head a little bit about, well, how do I kind of craft these things to drive some sort of end-to-end process? I need something to orchestrate them and I need something to discover them. So how does that all come together?

Absolutely, yes. And I think this is, this is also the way, this is how, um, we are actually trying to go do so the first and foremost, um, and in terms of is what we are doing is, you know, um, the, the way SSO we've actually positioned JUUL is to actually democratize access to business AI for every enterprise, right? So this is every user having, you know, their access to and in terms of is how they could interact with the business applications.

There's a new action bar that what we are providing that turns JUUL into an always on mode instead of users really invoking it so that um, no matter what the user is actually trying to go do they actually have a AI copilot that sits alongside with them and anticipate what's the user needs. And it is enabled to be able to provide the right recommendations for what the next step that the user should be performing depending on what their activity is. And our agents are actually functioning right behind the scenes for, based off of the activity that the users are trying to go do.

And they actually take either based off of a state, based off of, um, you know, in terms of as what the activity that the user is trying to go do or, or should perhaps possibly be. And they come to life performed by the jewel from an orchestration standpoint. And in terms of driving the end-to-end behavior, the thing that powers all of these things is also the way, this is how we have actually launched what we call as the SAP business wheat, something that we have announced as part of our unleashed event in February.

This is coming to life fully this time with the dual natively integrated. So what this is, is all of our application supporting end-to-end processes, whether you're supporting a source to pay process or you're in the process of doing a record to report, all activities are something that you can harmoniously e execute within the applications. The data underneath that is also harmonized and it's something that we are able to put to use in real time with the AI foundation that I talked about that is facilitated to be able to orchestrate on a continuous loop if you were to, and in terms of, and that's the true power of suite with this AI activated.

When I'm talk to people about AI agents, they're struggling with a, with one concept essentially, but a lot of business processes are deterministic, right? They're supposed to be done the same way every time and there's not a lot of room for variance. And with LLMs and AI agents, it's hard sometimes to get them to do the same thing the same way twice.

So how do I kind of marry probabilistic technologies with deterministic applications and how do I kind of stitch that together in a way where, uh, one augments the other? I mean, or how shall I think about this? No, this is really great question, right?

And I, I think this is indeed the challenge and in terms of is what everyone is facing and something that we're simplifying in a way is, is how we are providing this capability as well. So the first off is you, you talked about LLMs, right? So one of the things that what we are doing is we're abstracting the need for the users to having to know any of these large language models complexity.

We're abstracting that and that is something's very native behavior of the business ai. So that means depending on what the prompt is and what the activity that the user is trying to complete, the system determines the right set of behavior from utilizing the model and it latches to the appropriate model to be able to go in and do that. So we do the heavy lifting for the un, that's the fundamental power of business ai, right?

So now the second aspect of this is, you know, you, you're absolutely right in terms of any of these business processes are very deterministic. But that being said, you know, as you are going through any part, any particular process, like if you're in the process of a source to pay, right? I'll use this as an example.

Depending on what you are trying to go do, our agent as a first aspect of the agent is to be able to go determine based off of your past data as well as from a market insights perspective, do you have the right category strategy? And this continuously evaluates the market data and we actually have partnered with external data providers like Barrows and Moody's who can actually give you some level of data enrichment on top of your business data and helps you to be able to go to the right category strategies if you were to. So we interpret with the data to be able to provide the right set of things and that's what the power of our agents are able to go do.

And the business AI foundation abstracts these things, if you will. Now the category strategy said you need to actually go invest in this and based off of what your demand is, you need to go buy this. It is also intelligent enough to be able to go look at how is my supplier strength in some of these things that I intend to go buy based off of the demand.

And if the supplier strength is not good, it's intelligent enough to be able to go trigger the right sourcing event and that triggers the aspect of the sourcing that we need to go to. So that's the way SSO we're enabling the orchestration as well. So facilitating the user with the right behavior for the right next steps as well.

So hopefully that gives you the perspective. And in terms of, so we're abstracting the users from the complexity of not having to go manage any of these things. So our ways is how we interpret and utilize the prompts.

We abstract it and latch to the right model. And we also actually pay attention to, and in terms of as how we look at the data and state at that and in terms of is what it is facilitate the orchestration of the process. And this is not only limited to the way it is order the box supported SAP processes because we clearly understand every customer's business is unique and the customers can extend what we ship as the out of the box processes and it's geared to be able to work with extended process behaviors as well seamlessly.

So over time are the LMS becoming, for lack of a better phrase, disposable in the sense that I will invoke them through an agent depending on the capabilities required, the cost, and um, and, and frankly how smart they are because some of them have better reasoning capabilities than others, but sometimes I may not need that. And is that all gonna be managed and governed by you guys? So we actually partner with, um, every LLM provider in the market today.

So that is one of the fundamental thing that we strongly believe is something that we should go do. And we provide that so we hide the complexity of those from the end users standpoint so that the end users don't have to go manage whether you're actually doing an extended development or you are just a simple business user trying to actually go use the application. We hide that complexity from that and we take the burden and in terms of actually providing that, and that's, uh, something that we strongly believe in and hence the business AI Application Foundation, that what we are providing that paves the path for us people to support these behaviors.

And as these LLM behaviors changes as well, we train our data with across all of these things. And that's why it's a layer on top of any one LLM provider could possibly go in and do. We're able to go do that in a very unique way supporting our users.

What is the user experience gonna be like in the age of ag agentic ai? Because you know, for years we've been kind of navigating various graphical user interfaces, but maybe it's a much narrower experience where I start out with a chat with an agent and then it gets richer and richer depending on how I'm using these things. Or am I still looking at screens and then there's a bunch of agents that are kind of popping up all over the place.

It's a beautiful question and you know, there's three different ways this is how we are approaching this. So I think you're absolutely right. You know, the interface is morphing towards not the traditional persona based set of experiences is what we used to have.

We have completely reimagined that aspect as well. So it used to be where somebody would have to invoke a chat client. There again, we have announced, um, you know, as is what you might have noticed is with the new action bar, which is basically to actually turn JUUL into an always on mode.

So you have this action bar across every experience now that becomes a proactive ai co-pilot that anticipates user needs before they arise both in and out of SAP applications universe. And it actually gives you, uh, a much fundamental experience shift than what a traditional user used to in, in terms of, and that paves the path for us to be able to, depending on what the responses are, the prompts are to be able to trigger, um, the right agent. So that's one aspect of it.

The second flavor that what we are seeing is every organization is supported with more than 80% of casual users and about 20% of professional users. So we want to cater towards both of these sets of users. If you were to.

So we have what we call as a simplified set of user experiences, things like intake management, as an example, that facilitates asking our, providing the users with two to three sets of simple questions that they need to be answering based off of that it's able to actually go ahead and spawn the right set of activities behind the scenes to be able to go complete that task. So that's fundamentally is another way. SSO we are shifting the last but not the least is you're still not out from the perspective of these very rich set of professional user targeted set of experience.

We'll continue to have those as well, but we're pivoting more and more towards the first one, Mike, that I mentioned, which is the action bar type experiences where the user just starts with a simple bar. And this is also interesting and in terms of us, what we are trying to go to with the partnerships that what we're doing and one of the partnerships that I'll highlight is the perplexity that we've announced as well. So think of this as perplexity with SAP's action bar is no more just a simple search.

Now you have this infused with our data. So you have a very robust set of a B2B set of a search experience based off of your data right at your fingertips, no matter who you are as a persona within your organization. That's how we are re-imagining the experience into it.

And a very philosophical level. Our businesses have always been hampered by all the silos we have. Manufacturing, marketing, sales, um, finance.

Do you think in the age of agentic ai we might see entire enterprises and organizations get flatter in the sense that the, that I'll be able to navigate these silos a lot more easily using agents? I won't say, I mean, I I won't say flatter. I would probably say seamless in in terms of this how it is as is how we are seeing this.

And that statement is absolutely right and true and that's the way as is how we see this as well. And very simplistically said. And that is the goal of the business suite.

Uh, right, so that we are able to actually bring our power of, you know, something that we've always had, which is the unparalleled set of applications using the unmatched data in terms of something that we've been in the business for 50 plus years in terms of is what it is that we have been supporting with this business AI foundation. We're blurring the lines of all of these individual department silos as exactly as what you said. So that based off of data, based off of activity, we're able to trigger the right set, set of things completely blurring the seams of these sets of, um, departmental functions.

If you were to, so that it's more natural flows is how the process should get executed, you are able to actually understand and orchestrate accordingly. That's exactly as what we are doing. I'm loving, get your opinion on this 'cause it seems a little ironic to me, but historically when we see new technologies, it's always driven by startups and yet it feels like what gentech ai, the advantage might actually be towards the incumbent vendors that already have all the data that I need to drive the AI agents.

It is so fundamental that data plays a key role in all of these things. If you don't have the right data and an ability for you to be able to utilize the data in the best way possible. None of the steps about that, what we just discussed is going to be possible or feasible even for that matter.

And one of the true power of SAP is that we've been in the business solutions for over five decades and that truly paves the path for us to be able to put that data to use not just from an individual customer by customer perspective, but with that external set of vendors data that what we have been, uh, partnering with as well over the course of the last couple of decades. We have clearly have, uh, that as our key asset and that is the fundamental transformation opportunity for us to be able to put that to use. And without that data, nobody can actually go ahead and put these things to you, um, a better set of, uh, um, transformation behaviors.

And that's the challenge that every startup is facing because they don't have the underlying data, which is the true power that SAP has the preferring to be able to putting to use. And then that's, uh, something that you would see clearly, um, playing a key role. And that is also what is something that we have announced as with the Business Data Cloud, which brings all of the applications silos that are data very harmonized into one consistent place, which becomes a foundation for our business.

A and that's what is enabling the whole transformation, Mike And folks, you heard it here and the funny thing about AI is it begins and ends with the data. Hey Mano, thanks being on the show. Thank you so much, Mike.

Absolutely. And thank you all for watching the latest episode of the Techstrong AI series. You can find this in other episodes on our website.

We invite you to check them all out. Until then, we'll see you next time.