Europe's software giant fights to stay relevant in the AI era

Christian Klein takes a sweeping turn around his office building in southwestern Germany and hits the brakes of his black electric BMW. The chief executive officer of SAP SE jumps out in an ebullient mood, in spite of the circumstances. 

In recent years, artificial intelligence has rattled the enterprise software industry, raising doubts about the fundamentals of SAP's business model and leaving what was briefly the most valuable public firm in Europe racing to stay relevant. Members of the company's supervisory board think SAP needs an AI breakthrough in the coming months or it will fall painfully behind the competition, people familiar with the conversations told Bloomberg. While optimists believe that must come in the next two years, some directors believe one is more realistic.

"Of course, pressure is at its highest," Klein said in an interview at the company's headquarters in the city of Walldorf.

SAP built its success on pioneering applications for accounting and invoicing that gave customers access to all their data in one place while significantly speeding up processing times. It grew to become the world leader in enterprise-application software, and today, more than 90% of Fortune 500 companies manage their finances, supply chains or human resources through SAP tools. For half a century, Klein said, SAP engineers worked on an endless backlog of features that clients requested without ever having to rethink the underlying processes.

But new times demand new ways of working. Going forward, developers will have to sit down with clients to devise completely new AI-based methods, like automating supply chain management or financial statements as human involvement declines. "That's a huge, huge change," he said. "And that's exactly where we are right now — in the middle of it."

So far, it hasn't been easy. Three years into its AI rollout, which has largely failed to impress users, the 54-year-old firm has taken a beating in the stock market. Fears about AI replacing traditional software firms are pushing those companies — including SAP and its rivals, Oracle, Workday, ServiceNow and Salesforce — to compete in developing their own AI products.  

As no one company has pulled ahead of the pack, Klein rejects the idea that SAP has fallen behind. In the second half of the year, he's pledged to roll out 50 AI assistants that help users work faster, more than 400 AI agents to execute tasks autonomously and a new business AI platform. If all goes to plan, Klein said, "I can even see us leading the software sector." 

The coming months will demonstrate whether SAP can introduce AI tools at scale that clients actually want to use — or if the company starts to fade into the background as customers adopt other AI programs to run on SAP systems.

After intense corporate soul-searching, Klein has upended the company's management structure to put AI at the center of SAP's future. In March, he transferred oversight of the sales department to a colleague in order to focus on the new technology and beefed up a strategy he christened "All in on AI." Two months later, he presented his vision for SAP to investors, customers and staff at the company's most important sales event. 

Surrounded by the swamps and amusement parks of Orlando, Klein told the 13,000 attendees of the annual Sapphire conference to think of SAP as the "brain" for their enterprise, one that would allow users to develop and steer AI to work on its own. Independent AI agents could, for example, analyze millions of supply chain and sales data points to adjust financial forecasts. They "run the business and you can focus on what truly matters," Klein promised onstage.

While clients and partners told Bloomberg they found Klein's presentation compelling, what truly matters to them is whether he can deliver on it. "Christian did a good job of making the case about why SAP is more relevant than ever," said Geoff Scott, chief executive officer of Americas SAP Users Group, an independent body that represents the interests of SAP clients. "It's a promising world," he added, "but it is very much in its infancy."

Many SAP partners and customers agreed, saying the scenarios Klein described are years from becoming reality for them. 

One problem is that, after introducing its first set of AI tools in late 2023 under the brand name Joule, SAP has yet to produce any blockbuster AI products. Another is that many customers still use the company's legacy software, which runs on local servers. Until recently, SAP didn't even sell AI for its old systems. It initially took an all-or-nothing approach: offering AI access only to cloud users. The idea was to push more clients to adopt the new systems before the company cuts off regular support for its legacy software at the end of 2027.

"I can even see us leading the software sector."

Binding AI to cloud adoption proved to be a risky strategy. As SAP customers tend to be big businesses whose on-site software is heavily tailored to individual needs, switching systems is costly and complex. The stakes can be high if things go wrong. Last fall, medical tech company Zimmer Biomet sued Deloitte in New York, alleging it had lost tens of millions of dollars due to "business disruption" caused by the consultant's failure to properly migrate its old systems to SAP's cloud software. 

A spokesperson for Deloitte called the claims "meritless." A spokesperson for Zimmer Biomet declined to comment.

Even after years of planning and investment, there's no guarantee that these transitions will work as intended. This summer, German industrial company Zeiss Group — an important supplier for Dutch chipmachine maker ASML — backtracked six years into the process and after spending more than €200 million. Instead of building a new set-up from scratch as originally planned, Zeiss switched to reusing some SAP processes previously in place, and integrating them into a cloud IT set-up.

The migration "has now been realigned to achieve faster progress," a spokeswoman for the company said in a statement, declining to comment on costs or timeline. 

SAP's hardball approach of only selling AI for its new cloud platform also irritated clients, including some already in the process of migrating. "I think there was a feeling from the customer base over time that if I wasn't on SAP's latest platform, if I wasn't running the latest software, if I wasn't doing everything on the latest, latest, latest, there was no place for me," said ASUG's Scott.

The services SAP offers — like enabling companies to comply with financial regulations in more than 100 jurisdictions — can't be easily replaced, meaning the company is unlikely to see a mass client exodus. Moreover, no AI agents so far have been able to take on the tasks that SAP currently handles at scale. "We continue to see it as a pre-eminent system of record with a deep moat around its core business," UBS analysts wrote in late August. 

However, the analysts lowered expectations for SAP and its stock price target, saying that the valuation was "reasonable, but investor sentiment remains fragile." The company has been slow to roll out AI products, they added, which could increase "the prospect of customers choosing a DIY route to adopting AI in the interim."

In such a scenario, clients could take their AI money elsewhere, cutting SAP out of a potentially significant new source of revenue. Other companies would become partners for innovation. Threats like these have always haunted the industry: Siebel Systems, once the dominant player for customer relationship software, suffered such a fate after missing the cloud revolution and being overtaken by new players such as Salesforce. It was sold to Oracle in 2006.

Seven weeks after the Orlando conference, Klein implemented another top-level reshuffling. He stripped Chief Product and Engineering Officer Muhammad Alam, who's leaving the company early next year, of most responsibilities. Klein took on oversight of AI product development, and sent out a memo informing staff that all non-AI-related travel and hiring would be curbed in order to free up cash to invest in the tech. The moves were deemed crucial by industry experts. Significant investment is likely needed "to ensure SAP remains competitive," JP Morgan analysts wrote in reaction. Spending restrictions were also placed on internal AI use.

Those shake-ups have weighed on morale, especially since they've come after waves of job cuts, stricter performance management and a flawed new stock-compensation system. The latter caused an internal outcry when it was first rolled out in February and is now being reworked. 

In May, SAP's internal employee survey showed that trust in the executive board dropped to 54%. Earlier in Klein's CEO tenure, it had been at 80%. Excitement about the company's future also decreased "notably" by 8 percentage points to 62%, according to internal company communications reviewed by Bloomberg. While Klein maintained that his strategy for restructuring the company around AI "is clear, it has high approval," he acknowledged that "implementation is harder." With more than 100,000 employees across 81 countries, he said that making adjustments at this scale is a "huge" process.

Still, everybody agrees that changes need to be made, and as quickly as possible. When it comes to AI innovation, SAP is mostly selling aspirations rather than existing technology, several managers told Bloomberg, asking to not be identified as they were not authorized to speak to the media. 

According to several partners and customers, demand for the company's Joule AI products has been muted. An executive of a large pharmaceutical company said recently that his firm tried some of the tools, but ultimately decided against adopting them given how much work it would take to train those agents to focus on the right data. The executive asked to not be identified discussing sensitive business relationships.

Klein said in Walldorf that SAP's forthcoming AI-driven business platform is supposed to solve that problem by introducing context — or what the CEO calls "ontology" — which will enable agents to better grasp how a company operates. The platform, which will integrate 50 years of SAP's business experience, is now being rolled out and new functions are getting tested. "Hopefully, the feedback will reflect a significant improvement," Klein said.

"There will never be a solution that works out of the box."

SAP isn't alone in its struggles with AI. Many of its rivals are fighting to close the gap between clients' demand and what their AI tools are actually able to do. "The enterprise software industry has also been swept up in the AI wave," McKinsey senior partner Florian Bauer noted. Salesforce Inc., the leading developer of customer relationship management software, has launched its own AI platform, but can't always live up to its marketing. While Agentforce is often able to answer customer questions or collect information, the platform frequently falters when asked to take on more significant tasks. It turns out that while demoing AI might be easy, making the technology work within customers' complex IT systems is a whole other challenge. 

"Companies know where they want to go, but they aren't there yet," Bauer said.

In the meantime, some large SAP clients are bringing in outside armies of AI agents to automate tasks. Jason Gowans, the Chief Digital Officer for Levi's, told Bloomberg that it's a matter of finding the "right tool for the right job." "We don't think that there's one LLM model or one tool that can solve all our use cases," Gowans explained, referring to AI-based large language models. After nine years, Levi's migration to SAP's cloud software is set to wrap up in 2027, and for now, most of the company's agents run on Microsoft Copilot. Of the more than 1,000 agents that Levi's has already deployed, between five and ten are SAP's. He added that the ideal situation would be to steer all AI via Microsoft Teams.

Early this year, AI-induced fears led investors to offload traditional software stocks in what was dubbed the "SaaSpocalypse." The name referred to "software-as-a-service" companies like SAP, and mass sell-offs sent shockwaves through the sector. Analysts demanded that incumbents show progress in developing and selling AI, and in response, Klein ceded one of the key points of his strategy.

After hearing from several large customers that although it would take them several more years to migrate to SAP's cloud system, they wouldn't wait to start using AI, the CEO decided to enable AI integration with old, on-site systems. Klein described the move as "meeting our customers halfway," extending the offer only to large clients who agreed to shift at least half of the total value of their SAP licenses to the new cloud software.

This introduced a new set of technical challenges, and raised questions about whether the company was making promises it would be able to keep. Since SAP's AI was built for the cloud, two managers familiar with the matter told Bloomberg, it wasn't clear how — or if — it would operate on local installations. "There will never be a solution that works out of the box," Klein said.

In interviews at Sapphire, other SAP executives were tight-lipped. When asked about test cases in which AI was run both locally and in the cloud, two top managers refused to say if any already existed. 

Klein maintains that AI will be much easier to use when exclusively hosted in the cloud. To speed up transitions, he noted that the company was planning to launch seven migration agents by the end of the year, adding they would make the process 50% more efficient — which, if true, could save companies tens of millions of dollars. 

In Orlando, that estimate raised eyebrows among SAP partners and analysts. Some said 30% to 35% seemed more realistic. But even that could be a stretch. In May, analysts at the tech research and advisory firm Gartner cautioned that AI agents "aren't yet proven to cut expected migration costs or deliver the expected return on investment."

For now, customers are standing behind Klein and his big plans for AI. But many have made clear to him that their loyalty won't last forever. With plenty of other options available, SAP has to either join the new wave of AI giants, or risk being buried by them. 

When pressed in Walldorf on how much time he might still have to turn SAP around, Klein avoided a straightforward answer. None of the company's rivals have "cracked AI for enterprise," he noted. "I think we've all gone through a learning curve."