AI won't kill software — it will change your job, says SAP's Christian Klein

Key points

Key takeaways from Big Technology Podcast interview with SAP CEO Christian Klein (September 2026):

Business AI near-term. Klein expects per-task AI agents to reach 100% accuracy within months rather than years, with supply-chain optimisation agents following in roughly eight to nine months.

Accuracy by use case. Financial closing agents run at around 90-93% accuracy, which is insufficient for audited results; a 90-95% accurate customer brief from SAP's Joule assistant is good enough and removes the need for prep teams, so the required accuracy bar varies by task.

Data quality first. One New York customer runs about 100 finance systems, some from other vendors, and messy, siloed data cuts agent accuracy. SAP pairs its data platform with big data-lake providers for zero-copy sharing, while AI-assisted matching fixes data quality and reduces the need for armies of data scientists.

Faster planning cycles. Execution speeds up, shortening planning cycles to quarterly or monthly, so supply-chain, finance and workforce planning must now be built together with agent plans, since software and AI move much faster.

Change management, not burnout. Controls, compliance checks and approvals get automated, shifting people toward value-add work, and Klein reports no observed rise in fatigue among customers and end users, who are happy to shed approvals, but workload stays high, so heavy change management is needed.

Job security. Employees ask whether their jobs are secure, and Klein argues AI will not cause mass job loss: companies with large roadmaps and competitors to match do more, not less, using people plus technology, and the companies that survive will do more with people and tech.

AI governance. Because SAP runs mission-critical systems for governments and public-sector customers, agents must explain which data they access and share and where it is stored; governance and regulation are obstacles on what Klein frames as ending the cloud mountain and then climbing the AI mountain.

Read more: Big Technology Podcast