Where the Real Value Lies
Across industries, companies are pouring billions into AI, driven by real breakthroughs in capability and productivity. Yet many are struggling to translate experiments into measurable, enterprise-wide outcomes. The root causes are well-known: fragmented data landscapes, siloed processes, inconsistent governance, and AI bolted on to aging legacy systems.
Regardless of their industry or size, every customer I speak with wants one thing: AI that deeply understands their business and does so securely and reliably. That requires integrated applications, harmonized business data, and clear controls. Without these, AI operates in a vacuum, disconnected from business reality.
If it doesn’t understand how finance connects to procurement, how a supply chain interacts with manufacturing, what compliance rules govern a transaction, or how to handle exceptions, AI cannot reliably run a business. The smallest mistake – using outdated, incomplete, or incorrect data – can quietly cascade into wrong decisions, faulty transactions, and significant losses before anyone notices. Far from eliminating software, AI exposes the indispensability of the systems that coordinate work at scale.
Enterprise AI Succeeds Where Agents and Governance Meet
Building an agent is becoming increasingly easier – the tip of the iceberg. Deploying it across end-to-end supply chains or financial close processes, with full compliance and audit trails, is where most of the effort lies. Orchestration, policy enforcement, and workflow determinism are the gatekeepers of trust. The more autonomous agents you deploy, the more valuable the governed systems that constrain and supervise them become – and that’s where the platforms that already run the world’s core operations come into their own.
What Agents Need to Operate at Scale
To deliver real outcomes reliably, agents need three things. First, deep domain and industry knowledge encoded in systems, so agents understand context, relationships, and end-to-end processes. Second, accurate, semantically rich business data that provides a reliable source of truth. And third, enterprise-grade governance: validation rules, compliance checks, approval flows, identity management and audit trails to keep autonomy safe.
These are the elements that separate the AI that can truly and reliably run a business from the AI that merely impresses in a demo.
What Changes – What Stays True
AI makes software faster and cheaper to build. Large language models will be commoditized. Business models will evolve as usage patterns shift from users to agents. Entirely new interfaces will emerge. Users will increasingly converse with AI rather than navigate applications, and front-ends will be generated dynamically in real time.
But the need for continuously updated, governed systems only grows. AI raises the bar for secure updates, telemetry-driven improvement, and shared controls – all strengths of mature SaaS. AI agents don’t replace enterprise software. They rely on it.
The winners will not be those who own marginally better foundation models. They’ll be those who deliver value at the application layer: business outcomes grounded in deep domain expertise, integrated across functions, and governed for deployment at scale.
Software is becoming the operating system for trusted autonomy. The companies that recognize this will embed AI into the systems that run the world’s economy. The rest will run more experiments, generate more prototypes, and wonder why the outcomes lag the hype.
Long live software.
By Christian Klein, SAP’s CEO
