SAP AI customers link scale to governance, security, operational uses

For nearly four years, SAP has been riding high on the new wave of AI and releasing a relentless stream of AI platforms, tools and apps. Most recently, it has focused on agentic AI's need for context-aware information with a slew of data management platforms and partnerships.

But SAP's AI supply has yet to be met by strong demand from SAP customers. Early this year, Americas' SAP Users' Group (ASUG) conducted an online survey of 142 members whose organizations are engaged with AI in their SAP landscapes. While most were actively exploring SAP AI options, few had moved beyond pilots to implement AI across their enterprises. The survey, which was sponsored by SAP, Microsoft and Intel, also revealed that the biggest determinant of success was having the necessary governance, security, and operational alignment in place.

ASUG CEO Geoff Scott and Associate Research Director Blake Baltazar discussed the survey's takeaways, drilled down into some of the key findings and shared what they're hearing from members about their deployment challenges and SAP's AI strategy.

The interview was edited for length and clarity.

What are the takeaways from the survey?

Blake Baltazar

Blake Baltazar: First is that AI adoption is definitely advancing, but broader deployment and enterprise scale are the next hurdles, and customers will need a credible path from pilot to production. Second, how organizations approach AI matters more than just who they are. We've coined that the AI adoption mindset -- how members describe their organization's approach to AI. That reveals a lot more about their progress than their firmographics alone.

Intentional adopters are a bit more strategy-led and looking at business value. They're deploying AI in live processes much more than the measured adopters and are further along in their AI journey. They have executive involvement and are really looking at tying their priority AI use cases to the business. Measured adopters say, "Prove it to me before I scale." They're definitely more experimental and the bigger group of the two. They're looking more into building foundational knowledge and they take a more learning-oriented approach, testing and validating AI through pilots or experimentation before they get into deployment.

Even though they're at different stages, both groups have a lot to learn from each other.

The third takeaway is that execution readiness, not just the tech itself, is the primary barrier to adoption and scale. The most important barriers were governance, security and operational alignment.

Fourth is where people are using AI and seeing value. It's emerging from pragmatic and operational use cases. Our members are saying they're looking for AI to improve efficiencies first before they expand into more autonomous capabilities. They're prioritizing applications that are giving them an immediate operational boost.

Fifth is that AI ambitions are really clear. Organizations are widely expecting AI to help deliver on a lot of business outcomes, but the measurement frameworks are still catching up.

Geoff Scott

Geoff Scott: This conversation is at the heart of most of the dialogue we're having with SAP. The AI space is very fluent and fast-moving. When I talk to most of our members, it's like watching cars go by on the Autobahn at 150 miles an hour.

Our member organizations are risk-averse for the most part. You're taught not to just wing it with your SAP implementations. You're taught to be very careful and not make mistakes.

Many members are still in the middle of massive S/4HANA transformation programs and want to make sure they don't take their eye off that ball, and they finish those programs largely on time and on budget.

For them to look at this AI stuff, it needs to be more proven. Otherwise, they encumber significant risk in trying it out, having it not work, and then having a timing or scope delay.

This fall, ASUG is giving members ways to use AI to accelerate their S/4 transformations. Many members see them as two bell curves that happen to be in sequential order with each other. While I respect that most IT organizations say they want to do S/4 before jumping into these other things, maybe their line-of-business peers, CEOs and boards have slightly different perspectives. They'll need to figure out how to juggle both, and that's a really big deal. Most of our member organizations are working 40-plus-hour weeks just getting S/4 done.

SAP has been trying to move people to the public-cloud, multi-tenant SaaS version of S/4HANA. How much of your members' AI efforts depend on cloud deployment? You don't necessarily have to have everything in the cloud -- even AI.

Scott: As an SAP customer, you can largely do whatever you want as long as you expend the energy and effort. If you're an [ERP Central Component] customer and absolutely focused on creating AI capability inside your ECC environment, do it. However, those are not going to be SAP-sponsored solutions.

But when there's a will, there's always a way. The question is, do you have the will? If you have the will, are you also willing to spend time to keep it going? How much of this estate do we want to continue managing, and how much do we want to give back to SAP to manage on our behalf?

Gone are the days when we can take parts of SAP and customize them to our heart's content. We've bent it into what we want it to be. My concern is that bending, over time, becomes unmanageable and unmaintainable. If you really want to get the most out of your SAP landscape, the very best thing you could do is drive toward public cloud, where you're giving back SAP functionality you've taken on your own to run and looked at it very hard and said, “This isn't worth it anymore.”

But in doing so, you have a big hurdle to climb with line-of-business peers who say, "This is the way we've always done it. Why would you make us do it any other way?" One of the things that we've been saying is: Run towards public cloud. That necessitates SAP having public cloud solutions that customers can actually adopt. Then it frees them up to do the more important work, which is innovation. It's really hard to innovate and adapt in a fast-changing landscape if you're encumbered by all the maintenance you have to do every day just to keep the lights on.

Were there any surprises in the survey?

Baltazar: We asked, what would increase your confidence to move forward? What do you need to either adopt or expand AI? We found -- and this is consistent across all the adoption mindsets -- that practical, specific guidance for IT, finance, operations and developers was noted as the top need. Right after that, at 51%, was step-by-step implementation support tied to SAP solutions.

What surprised me was that what gives us confidence is the same. No matter where we're at in the AI journey or how our organization prioritizes or what motivates us to use AI, the things that will increase our confidence to move forward with it are the same.

Scott: Two things surprised me. First – this is the good news – is that member organizations are playing around with this, and I'd love to see a lot more experimentation. The word "experimentation" is important because it also says, "I'm going to try something, and it's okay if it doesn't work. I'm going to use that learning to reset and try again."

For a lot of us in the SAP space, that's unfamiliar territory. We're not terribly used to things not working. We're used to banging at them until they do.

The second surprise is that challenges with data-quality readiness only come in at 25% as a barrier. Maybe the reason is if I'm using AI to provide support, guidance, documentation -- what I would consider supporting activities -- if I'm going to realize a future agentic world where agents make more specific recommendations and take over parts of a business process, the data quality issue is going to rear its head and become a much bigger issue. It might become the biggest single gating factor in why agentic is not adopted in an SAP ecosystem -- because the data inside the SAP estate isn't clean enough or reliable enough for early-stage agents to make the right decisions.

I think we will solve this over time as the agents get more sophisticated and we get further along the curve. But on day one, it's going to be a lot of blind faith. It's like the first time you take your teenager up on the highway and they're behind the wheel of the car, and you're sitting in the passenger seat praying to God they understand how to brake and steer. These agents are going to be in a very similar model. If the data's not right, they're going to spin out all over the place. Do you want that in front of your customers?

The S/4HANA transformation that we've been discussing has been lift and shift. I'm going to pick up my ECC estate and make some changes to it, but I'm going to narrow those changes because I don't want this to take forever. And I'm going to drop it into S/4.

As I have talked to many of our customers and we've had this debate internally, some of that data quality and data readiness work has not been done.

We may get to a point where you're ready to turn all this on and find out the data is not where it needs to be for the training wheels to come off. The road has a ton of potholes.

We may get to a point where you're ready to turn all this on and find out the data is not where it needs to be for the training wheels to come off.

Data has been a big focus of SAP. Are the SAP tools and partnerships rising to the occasion?

Scott: We're starting to see some of this. Some of the partner firms are doing more work, but I think a lot of that work has been done in deference to the transformation effort -- getting people from point A to point B.

We started S/4 transformation 10 years ago. This AI thing did not exist, so there wasn't any pressure on our organizations to look at data. Now, SAP is outstanding when it comes to creating structure around data. As customers, we've been able to create a lot of structure and inside that structure insert wonky or wobbly data -- even to the extent that how I ran my business two years ago is not how I want to run it today or how I ran it 10 years ago. All of that exists inside that SAP estate, including all those legacy things that may or may not be especially helpful when I want to think about a future agentic estate.

Think about it this way. SAP is an amazing apartment building. They have built an apartment building everyone wants to move into, but inside those apartments, people have decorated them however they see fit. Some apartments could look super-modern, some super-conservative, some could look like someone backed up a truck and just dumped stuff in the living room.

Some of this leads to massive conversations inside the four walls of the customer. How exactly do we want to govern master data? What is the definition of a customer or a product?

SAP will support you in a myriad of ways. But if you're not consistent in that application, and then you turn on an agent, and you say, "Agent, we've been all over the place in how we've run this. I can't figure it out” -- are you going to expect the agent to discern all of that?

There may be a lot of work that our customers need to do to get their data house in order before they're willing to turn on agentic tools. We're making a giant leap in wishing these tools will figure out how to do it when we haven't.

Scaling AI across the enterprise seems to be the number one topic for partners, consulting firms and system integrators. What did the survey tell you about progress in scaling AI?

Baltazar: There were security, privacy and governance concerns. There were also skills gaps, budget constraints and, of course, data-readiness challenges. About 10% say they're at this stage. They need to know how AI applies to specific teams, functions, and roles across their business before they can feel confident enough to scale enterprise-wide.

Scott: The most important thing any enterprise can do right now is experiment with the hope of scaling. When you look at the size and scope of enterprises that run SAP, scaling is going to be a significant undertaking.

But I don't think you use that as an excuse to not get going. You use it as an enabler to figure it out. The first thing you need before you can even do that is a team that's knowledgeable about this. And the only way they get knowledgeable is through experimentation and learning.

David Essex is an industry editor who creates in-depth content on enterprise applications, emerging technology and market trends for several Informa TechTarget websites.