SAP is expanding its AI models for business data with tools for anomaly detection, forecasting and automated decisions. The practical shift is from predicting what might happen to helping businesses act on structured records such as orders, invoices and supply-chain exceptions.
SAP Watch analysis
What happened
At its TechEd conference, SAP introduced SAP-RPT-2, which adds anomaly and outlier detection, creates structured data embeddings to compare similar records, and can analyse information across related tables without first combining them into one. The AI Economy reports that general availability is planned for the first quarter of 2027.
SAP also introduced two specialised models. SAP-RPT-TS forecasts changing business measures such as product demand and is available now through SAP HANA Cloud or SAP AI Core. SAP-RPT-Decisions is designed to automate choices inside applications and agent workflows, including approving customer orders or routing supply-chain exceptions; SAP expects it to be generally available through SAP AI Core in the fourth quarter of 2026.
Why it matters
These models target the structured data and routine decisions already embedded in enterprise systems, rather than asking a general-purpose chatbot to improvise its way through a spreadsheet. Spotting unusual transactions, forecasting demand and routing exceptions could give businesses more ways to bring AI into operational workflows.
The timing differs: SAP-RPT-TS is available now, while the other two models have future availability dates. That makes the forecasting tool the one customers can evaluate today; the broader decision-making pitch still has to become a delivered product.
Our read
The interesting move is SAP’s focus on the unglamorous data behind business operations. An AI that can work across related tables and help route an exception may be more useful to some companies than another fluent chatbot. But the announcement is a product roadmap, not evidence that these models improve accuracy or business outcomes. The proof will be whether customers can use them reliably in real workflows, without adding a new layer of complexity to the old ones.
What to watch
- Whether SAP-RPT-2 arrives on schedule in Q1 2027.
- How SAP measures accuracy for anomaly detection and automated decisions.
- Whether SAP-RPT-Decisions becomes generally available in Q4 2026.
- What customers can do with SAP-RPT-TS in practice.
Discussion spark: When AI moves from forecasting business outcomes to making operational decisions, which should companies demand first: measurable accuracy or clear human oversight?
Sources and evidence
- SAP Teaches Its Tabular AI to Make Business Decisions, Not Just Predict Them – The AI Economy | Ken Yeung (27 October 2026, 08:02 UTC)
not affiliated with or endorsed by SAP