SAP CEO: the AI race is being fought in the wrong place

The article argues that enterprise AI value will come from embedding models into operational systems, not from standalone copilots or prompts. It emphasizes context, governance, process integrity, and transactional understanding as the key enablers of execution, risk reduction, and coordination across functions. The piece is strategic commentary rather than company-specific news, so immediate market impact is limited.

Analysis

The investable shift is not generic AI adoption; it is a re-pricing of systems that own workflows, permissions, and transaction data. That favors incumbents with deep ERP, SCM, CRM, and governance embedment because they can turn AI from a chat layer into an execution layer. In the next 12-24 months, the market is likely to reward vendors that can prove closed-loop automation metrics — lower cycle time, fewer approvals, lower error rates — not just higher seat counts or copilots shipped.

Second-order, this is more defensive than the market assumes for the infrastructure layer and more mixed for the app layer. Model providers can still win usage, but capture may be limited if the control point sits inside enterprise software where the context lives. The biggest losers are point-solution startups built on thin integrations: once buyers demand compliance, auditability, and dependency-aware execution, those products become features, not platforms.

The contrarian view is that the near-term market may be overpaying for “agentic” narratives while underestimating implementation friction. Real autonomy increases deployment friction because firms must reconcile permissions, data quality, exception handling, and human accountability before scaling. That means revenue from AI transformation may arrive slower than hype suggests, but once embedded, switching costs and gross retention should improve materially for the winners.

Catalyst-wise, expect a wave of budget reallocation over the next 2-4 quarters from experimental AI spend toward workflow-native vendors that can show hard ROI in finance, procurement, customer operations, and supply chain. The key risk is a pullback if buyers conclude copilots do not materially reduce headcount or working capital; however, a more likely outcome is selective adoption in high-friction processes first, with measurable productivity gains showing up gradually rather than all at once.

Key Decisions for Investors

• Long MSFT / short a basket of thinly integrated AI workflow startups via public comps or venture proxy exposure; thesis is that control-point software captures the AI budget once enterprises require permissions, audit, and transactional context. Time horizon: 6-18 months; target is multiple expansion for incumbents versus compression for feature-level vendors.
• Long ORCL or SAP on a 3-6 month horizon into enterprise budget season; risk/reward improves if management emphasizes AI attached to ERP/SCM execution rather than standalone copilots. Use pullbacks on any AI hype selloff as entry, with downside limited by recurring-revenue durability and upside from higher attach rates.
• Pair long SNOW / short a basket of generic LLM-exposed software names if you want to express that data context becomes more valuable than raw model access. The edge is that context ownership should drive monetization as enterprises move from experimentation to deployment.
• Buy 12-18 month call spreads on NOW; it sits closest to the workflow-to-execution conversion point, and the market may still underprice the second-order lift from AI increasing case volume and automation depth rather than just lowering service costs.
• If you want a contrarian hedge, short a basket of pure-play agent/orchestration names on any sharp rally and cover into product-announcement spikes; the risk is narrative-driven multiple expansion, but the reward is that implementation complexity should delay revenue realization and lead to post-hype compression.