Share with your CIO
SAP is betting that AI agents embedded directly into core enterprise workflows are the company’s next major growth chapter, and CEO Christian Klein is making that case publicly while the company repurchases its own stock at pace. Between September 21 and 25, 2026, SAP bought back 590,000 shares on Xetra, bringing the program’s cumulative total to 8.9 million shares. New product moves include TabPFN-3.5 Plus for in-platform forecasting, plus extensions across Signavio, LeanIX, and Cloud ALM. Customer deployments span Lockheed Martin on HR, Team Liquid on performance analytics, and BERNMOBIL on cloud modernization. Q3 results land October 21.
What this means for your business
If your organization runs SAP as its ERP backbone, the AI agent pitch isn’t a roadmap abstraction anymore. Klein is describing agents woven into data structures and operational steering, which means the integration points your team owns today are the same ones SAP intends to instrument with autonomous decision-making. Whether you’re a net beneficiary or a net risk depends almost entirely on how cleanly your data estate sits inside the SAP perimeter, because agents running on dirty or fragmented data create faster wrong answers, not better ones.
The product releases this week follow a pattern SAP has run before: announce the vision at the CEO level, then surface product-level evidence to show the vision has traction. TabPFN-3.5 Plus is a legitimate signal, a machine-learning model capable of rapid tabular forecasting without extensive training data, now available natively inside SAP AI Core. That matters because it shortens the time between “we have a forecasting problem” and “we have a running model” without requiring customers to provision separate infrastructure. The risk is that SAP’s track record on AI feature depth has lagged its AI narrative depth, and October 21 earnings will be the first hard read on whether cloud revenue is actually accelerating or whether the agent story is running ahead of the bookings.
The buyback is a confidence signal worth reading carefully, not as proof of health but as a specific claim about capital priorities. A company uncertain about its AI investment cycle doesn’t typically return cash at this rate; it hoards it. That’s either evidence that Klein genuinely believes the AI product portfolio is self-funding its own growth, or it’s financial engineering designed to prop earnings-per-share while top-line growth is still transitioning from licensed software to cloud. I’d revisit that read if Q3 cloud revenue growth comes in below the double-digit rate Jefferies and Citi are implicitly pricing in with their upgraded targets.
Concept deep-dive: AI agents in ERP
An AI agent, in the SAP context, is software that monitors a business process, decides when a condition warrants action, and executes that action without waiting for a human to click approve, think of it as a junior analyst permanently watching one dashboard and empowered to act on what it sees. Inside an ERP system like SAP, where procurement, finance, HR, and supply chain data converge, an agent can trigger a purchase order, flag a compliance exception, or rebalance a workforce plan. The business consequence is that process latency collapses, and so does the margin for data quality error.
Based on reporting from SAP Keeps Foot on Buyback Pedal as Klein Pitches AI Agents to Enterprise Customers, originally published 2026-10-03 10:51:00.
