SAP’s Connect conference, held in Las Vegas in early October, was the first major touchpoint between SAP, customers, and partners following SAP Sapphire & ASUG Annual Conference. As such it provided an opportunity to review SAP’s progress towards the strategic goals set forth in Orlando, as well offer a sanity check on how those goals have been resonating with customers since May. 

As expected, AI, agents, and the Autonomous Enterprise were front and center in most of Connect’s keynotes and conversations, even as it was clear that customers are more focused on laying the groundwork for this vision than building it outright. Nonetheless, SAP’s description of an emergent focus on agent security and governance at Connect was a welcome refinement to its overall AI strategy, considering the unsettling spate of news stories about rogue agents run amok. Also well-received were the additional details SAP provided regarding  its ability to add significant business context to its LLM-based AI models and the agents that serve as their user interface. 

A welcome degree of pragmatism was also evident. Despite the AI-first rhetoric, underlying the conference was a recognition that customers need to be met where they are on their journey to the Shangri-la of enterprise autonomy. SAP leaders with whom I spoke also acknowledged that SAP needs to produce a critical mass of customer proof points in order to show the majority of the customer base — who are still working through the fundamentals of their cloud-based, S/4HANA migrations – the way to an agentic AI future. 

But achieving the main goal of Connect remained elusive in 2026, as was the case with the inaugural Connect last year. Connect’s original raison‑d’etre, as envisioned by outgoing SAP board member Muhammad Alam and others, was to be a platform for showcasing how SAP’s product and technology goals resonate both within the five principal lines of business represented in this conference — finance, supply chain, human resources, procurement, and customer experience — as well as between these lines of business. 

In that regard Connect 2026 fell short. A combined within and between message was brought home in the main keynote, which also happened to be one of the best keynotes in recent memory. But the rest of the conference largely defaulted to messaging around individual lines of business, with a paucity of moments when, either from an LoB keynote stage or within a session, a speaker talked about why busting silos and building end-to-end processes—connecting the LOBs represented in the conference — should be an important part of their overall business strategy, and their AI-specific strategy as well. 

While the message behind Connect’s original intent fell short, the other key messages noted above landed relatively well, albeit more in theory than reality. The security and governance message, delivered during the main keynote by Marielle Ehrmann, SAP Chief Security Officer, and one of the very few senior female leaders on any keynote stage, was notable for both its comprehensiveness and the enormous challenge that fulfilling its promise will represent for SAP. The promise Ehrmann put forth, based in part on Signavio’s agent mining capabilities, involves both setting appropriate boundaries for what agents are and are not allowed to do as well as monitoring agents’ behavior to ensure they stay within those boundaries. 

It’s an excellent and timely vision, and I have no doubt that, from a pure technology standpoint, SAP can build these guardrails and deliver them to customers. But the scale of the monitoring task seems particularly daunting, and clearly more details of how this capability will unfold are needed. 

It’s the scale of the problem SAP will face when large numbers of customers put their agentic armies to work that’s most troubling. SAP isn’t just interested in managing its own growing agent portfolio. Multiple keynotes made it clear that SAP will support agents developed by partners and customers using Joule, as well as agents built using other agentic frameworks. This platform of platforms approach is one of many ways in which SAP wants to usurp or corral the encroachment of rivals such as Palantir and others in its customer base. 

But the potential scale of the monitoring requirements seems astronomically complex: SAP told me that its own experience as agentic AI customer zero has recently surpassed one million agentic interactions to be monitored per day. That’s SAP, which can tightly control the usage, quality, and scope of these agents across a highly-modernized backend that is well-bounded and, compared to many of its customers, based on a relatively straightforward business model and supply chain. 

What will the scale of this effort look like when ten, or twenty, or two hundred of its top customers need this level of agentic monitoring? What happens when there are thousands, many using sovereign clouds to run their businesses? What will the effort look like as the non-SAP and non-Joule agents are added? As I just said — more details are clearly needed. 

And soon. The business context focus behind SAP’s agent strategy seemed to be making more and more sense, and while most customers are still working on getting the foundation for an eventual AI strategy in place, SAP’s case for why it is the agentic partner of choice for enterprises that already run on SAP seems to be stronger each time they tell it. 

Another opportunity that needs to catch up to reality quickly could be seen in a promise from CEO Christian Klein. Kicking off the Day One keynote, Klein promised the end of ​“three year migrations” for its customers. And underlying this promise was SAP’s toolchain— Signavio, LeanIX, CloudALM, WalkMe, in addition to partner products — and an agentic layer that will make using these tools and associated methodologies like Activate significantly easier and more effective. 

But it’s not enough to just have the products and tools, they have to be widely adopted as well. Right now, the vast majority of SAP’s existing customers have a huge backlog of ECC and older S/4HANA instances they need to migrate and transform, and the vast majority of these customers are not using the toolchain in the integrated way that will lend itself to fulfilling Klein’s ​“no more three-year migrations” promise. This is particularly true in North America, where Cloud ALM has singularly failed to take hold in the way SAP had hoped. 

In other words, another solid strategy in need of some serious execution in 2027. 

And one that’s an important linchpin to fulfilling the rest of the strategies we heard at Connect. Here’s why: SAP’s ability to add business context to its AI models and agentic workforce, based on 30-plus years of delivering business value to customers, has the potential to put SAP head and shoulders above its AI rivals. This deep business context, when deployed at the design and build phase of a project, also allows SAP and partners to match business requirements to technology solutions in ways that no other competitor can potentially match. If SAP can pull off the agentic security and governance strategy it highlighted at Connect, and make the case that achieving this marriage of business requirements and technology is possible without suffering through a three-year implementation, then SAP will truly be the company to beat in AI and enterprise software.

But none of this can happen if the toolchain and its partner products continue to languish on the sidelines as techy tools for techies, instead of becoming business tools for business stakeholders as well. To do this will require a concerted effort to showcase the breadth and depth of this opportunity, and one of the best ways to do that is for SAP to realize the unfulfilled promise of Connect as the place to tell this silo-busting, end-to-end process transformation story.

If SAP could not only shout those end-to-end process stories from the rooftops instead of whisper them from the wings, but also show how the toolchain can empower line of business stakeholders to see, understand, and get behind the value that a truly connected enterprise can provide, the uptake problem for the toolchain could potentially be solved. And if every customer looking to transform their company to a modern, clean core, integrated SaaS-based enterprise was signed on to using the toolchain, not only would the ​“no more three year implementations” promise be realistic, the resulting net-new, integrated process landscapes, built with LoB stakeholder buy-in, could become the perfect staging ground for whatever version of an agentic AI future the customers would like. On their terms. 

This would require resurrecting the so-far failed original promise of Connect, and it’s not clear if they can. My fear is that the process silos SAP wants to break down in their customer base are too well reinforced by product silos inside SAP. It’s back to what I talked about in a long LinkedIn piece I published recently: cleaning up process silos — and messy data — requires a level of change management most SAP customers have little experience with. It may be that the mother ship has a similar change management problem as well. 

Or does it? 

Looking for more information on Connect? Here’s some links that might be useful for you: 

• Jon Reed, Bonnie Tinder, and Josh Greenbaum did a watch party for the Connect keynote. You can watch it here: SAP Connect Live Watch Party — 2^nd annual — Join us | LinkedIn
• Jon Reed’s analysis: SAP Connect 2026 — SAP leadership responds to hot AI issues, and explains why ​“generic models” fall short
• ASUG’s own Jim Lichtenwalter, Kelly Dowling and Elizabeth Tuckwell discussed what they saw at Connect on the ASUG Talks podcast: Key Takeaways from SAP Connect 2026 | ASUG

Joshua Greenbaum is the Principal at Enterprise Applications Consulting