No PriorsSAP: Bringing the ‘Operating System’ of a Company into the AI Era with CTO Philipp Herzig
At a glance
WHAT IT’S REALLY ABOUT
SAP’s CTO on scaling enterprise AI, agents, and outcomes-based software
- SAP positions itself as the “operating system” for enterprises, spanning finance, HR, supply chain, and customer-facing workflows across 400,000 customers.
- Herzig argues SAP’s durability comes from standardizing repeatable business needs while continuously re-engineering for major platform shifts (mainframe→client/server→internet→cloud→AI).
- SAP’s AI transformation targets three layers at once: generative/proactive UI, agent-driven process execution (“outcome as a service”), and a harmonized semantic data layer to ground AI in enterprise truth.
- The hardest enterprise AI problem is not demos but scaling reliable, contextual behavior across huge document corpora, complex master data, and tens of thousands of APIs with strong security guarantees.
- Beyond LLMs, SAP is investing in specialized predictive/tabular modeling (RPT-1) to democratize forecasting and decision support that classical ML can do but doesn’t scale organizationally today.
IDEAS WORTH REMEMBERING
5 ideasEnterprise AI success is an outcome race, not an innovation race.
Herzig frames a growing gap between flashy AI innovation and measurable enterprise outcomes; SAP’s strategy emphasizes reducing effort, time, and cost in real workflows rather than standalone demos.
Scaling matters more than the initial prototype.
RAG and MCP-style integrations look easy on 10 documents or 10 APIs, but SAP customers require personalization, policy correctness, and orchestration across thousands of documents and ~20,000 APIs.
AI is forcing a redesign of the user interface model.
He predicts the end of “UI that teaches humans to click” and a move to generative, proactive, multimodal interfaces that surface issues (e.g., supply chain disruptions) and propose actions.
Agents will shift software from SaaS to “service/outcome as software.”
Instead of rigid end-to-end processes, agents blend structured and unstructured work—handling documents, exceptions, and coordination—while humans supervise higher-level decisions.
Verifiability requires enterprise-grade evals and boundary conditions.
Code agents work because compilation/tests verify outputs; for finance/HR outcomes, teams must encode expected outputs, constraints, privacy/security rules, and ongoing evals to ensure reliability.
WORDS WORTH SAVING
5 quotesSAP is… kind of the operating system… of a company essentially.
— Philipp Herzig
The time is clearly over where you design software… that requires the intelligence to sit in front of the computer.
— Philipp Herzig
The biggest challenge… is… teaching the AI to do the right thing at scale.
— Philipp Herzig
In the past, we called this process mining, now we call it agent mining.
— Philipp Herzig
Our job at SAP is to make the technology disappear.
— Philipp Herzig
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome