Late-20th-century history books generally view Mikhail Gorbachev as the man who brought about the dissolution of the Soviet Union by trying to save it through reform.
One major reform, glasnost, or openness, mostly worked, but perestroika, or restructuring, is widely seen as an ill-conceived rush to change that led to economic ruin and an eventual return to a closed society.
The analogy was on this history buff's mind last week when SAP used its annual Sapphire conference in Orlando to advance its strategy for providing customers with the data infrastructure they need to deploy AI successfully on its platforms. SAP announced numerous new products and updates, promising to ship them by the end of the year.
Facing pressure to stay atop the AI wave and not get swamped by it, the vendor finds itself having to offer an unprecedented degree of openness to the vast stores of business data customers have collected on SAP ERP systems over the 53 years it has existed.
That data, along with the well-honed business processes associated with it, might be SAP's biggest advantage in the age of AI. SAP executives certainly made the case at the conference, often characterizing it as SAP's bulwark for weathering the AI winds. But since AI often needs access to non-SAP data, the information and know-how captured in SAP ERP systems must be part of a much bigger pool of data than in the days when ERP was a mostly closed-off, monolithic system.
All this puts SAP in the challenging position of simultaneously having to defend the crown jewels while making them more accessible. Navigating these waters will be tricky because it comes at a time when AI is also forcing structural change on SAP, with other players -- including numerous partners -- competing in agentic AI applications or jockeying to be the AI-assisted portal into enterprise applications, while others try to be the preferred solution to the data-access problem.
Jon Reed
"This is the biggest shift for SAP. This is way bigger than the internet," said Jon Reed, co-founder of Diginomica.
Open for competition
Interviews with SAP executives at the show and with industry analysts shortly after paint a picture of an impressive, multi-front effort to respond to the AI moment by exploiting the unique advantage of SAP ERP systems, which SAP never tires of saying carry roughly 80% of the world's commercial transactions.
SAP's "functional stack" is indeed powerful, said Joshua Greenbaum, principal at Enterprise Applications Consulting, "when you look at the fact that SAP brings the data, the business concepts, the process, the history -- decades of work deep in 26 industries. They provide compliance as a service, regulatory management as a service. They do these things that are a functional and infrastructure stack layered underneath the business process functionality, layered underneath an agentic or assistant layer."
Building on top of SAP ERP and taking advantage of its built-in governance and security gives developers an advantage in solving the data challenges, while not incidentally confirming SAP's indispensability.
Indeed, demonstrating that partners are innovating on its platforms will be critical to SAP's strategy, according to Reed.
"One of the key things vendors like Oracle, SAP and Salesforce have to do is encourage their partners to disrupt them a little bit," he said. "You would much rather your partners take some of your business with agentic applications than your non-partners." The 100-million-euro investment in AI partners SAP announced at the conference -- to great applause -- shows it is aware of the need to bolster the partner ecosystem, he said.
Holger Mueller
But SAP has made missteps in this area before, according to Holger Mueller, vice president and principal analyst at Constellation Research. In an email discussion, Mueller recalled a time when SAP had no mobile applications of its own, and then-CEO Bill McDermott ordered a rush job. "They were never useful," he wrote.
SAP's new AI agents were built in a similar way, Mueller said, with developers flown to headquarters in Walldorf, Germany, and offices in Palo Alto, Calif., and holed up for two weeks until agents were built. "The future will tell how good they are," he said.
AI-driven disruption might have positive effects, but it also opens SAP to competitive threats from several layers of the AI architecture. For example, vendors of integration, workflow and business process management software, such as Boomi, ServiceNow and UiPath, are moving toward providing the governance and orchestration layers for agents, Reed said. Google wins many of the agent orchestration purchases, while Anthropic's Claude Cowork agent is gaining ground thanks to its UX for enterprise applications.
"Everyone's trying to make this play where they can manage everyone else's stuff," he said.
SAP and top ERP competitors, including Oracle and Workday, have thought about these scenarios, Reed said. "That doesn't mean they're left out necessarily, because they can still make the case to Anthropic that they should still be calling SAP or Workday or Oracle agents when that kind of information is needed. They'll say you're going to get a much better result than just trying to pull our data on your own."
Business Data Cloud: SAP's solution to AI data chaos
The move toward openness is forcing ERP vendors to play nicer with others, according to Reed. "SAP clearly is struggling at times to come across as a totally open vendor," he said. "Historically, it hasn't been perceived that way."
Mixed messages aren't helping. Three weeks before Sapphire, SAP published a new API policy restricting third-party autonomous agents from accessing SAP data outside SAP-approved channels.
"Customers are going to get really pissed off when you start setting up data toll roads and saying you can't connect with this or that," Reed said. On the other hand, he said, all of SAP's recent data-related acquisitions show it is trying to be more open.
Irfan Khan, president and chief product officer of SAP data and analytics, said SAP's internal effort to incorporate data from platforms it acquired, such as Ariba, Concur and SuccessFactors, is exactly what prepared it for the third-party data demands of the AI age.
"We had a very sophisticated data model that was supporting multiple industries," Khan said, alluding to the company's first four decades of organic growth, during which it built a substantial ERP heritage. "Then in the last decade, we went on this massive spending spree and probably ended up accumulating $55-$60 billion worth of SaaS companies."
SAP developed its primary data platform, Business Data Cloud (BDC), to harmonize the data models of SAP ERP and the SaaS applications, Kahn said. One key component is SAP data products: prepackaged data assets that combine data, algorithms and metadata to make common business records, such as sales orders or supplier data in procurement applications, shareable through BDC.
Another big advance in SAP's effort to make ERP and third-party data available to AI is zero-copy sharing, a recently added feature of BDC. "It means we don't physically have to copy the data to any one of these environments; we just put it into the BDC environment and share the data," Khan said. "Like with a SharePoint server, you allow access to the data from any constituent environment, like a Snowflake or Databricks, as if it was local to them."
For Mueller, SAP's announcement that it is moving the cloud version of its in-memory database technology, HANA, into BDC is a significant development and could solve SAP's remaining data problem. "Agents need lightweight transactional databases to record data. HANA is not lightweight, but it is what SAP has." He added that HANA can replicate data easily and out of the box, enable SAP to move its transactional data into BDC and relieve the vendor of the need to build new data products.
Word power: The autonomous enterprise
While it was unveiling an onslaught of products at Sapphire, SAP took the opportunity to rebrand its strategy as a way for customers to build the "autonomous enterprise." Analysts questioned the wisdom of making that the brand for SAP's AI pitch, including HR technology analyst Josh Bersin, who called it "not the best message" in a post that nonetheless lauded, and described in detail, the architecture SAP is building to carry out the vision.
Even if it's possible to build an autonomous enterprise, who would want one?
Reed said it remains to be seen how much SAP customers will be motivated by the phrase, which he said has probably done more to energize SAP itself. Customers want to use AI to automate at their own pace and in processes that make the most sense to them, and they're doing so amid employee concerns about being replaced by technology.
Joshua Greenbaum
While the analysts all found ways to compliment SAP for providing practical tools for deploying AI, Greenbaum said it offered mostly theoretical use cases at Sapphire and few case studies of customers who are actually getting value from it. He noted that in many cases, the preparatory work of integrating workflows for AI will be what produces positive ROI, not the AI itself.
"How do you separate the AI value from the basic transformation value? It's going to be a big, big question," he said.
David Essex is an industry editor who creates in-depth content on enterprise applications, emerging technology and market trends for several Informa TechTarget websites.
