SAP’s Autonomous Enterprise Vision Will Test More Than Technology: A Conversation with KPMG Leaders

SAP Sapphire 2026 gave SAP customers and partners a new North Star: the ‘autonomous enterprise.’

For KPMG SAP professionals, the significance is not only that SAP is positioning Joule and the SAP Business AI Platform as the next phase of enterprise software. It’s also about how customers will decide where autonomy belongs, where humans remain accountable, how agents are governed, and what operating model is needed to move from pilot to production.

In a conversation with SAPinsider, Michael Golz (left), CTO for SAP Advisory, KPMG US and Valentino Koester (right), Global Head of the SAP360 Program and Head of SAP AI & Data, KPMG International, discussed SAP’s autonomous enterprise strategy, the readiness gaps they see with customers, and why governance should be embedded into agentic programs from the start.

Q: SAP Sapphire 2026 put the autonomous enterprise front and center. From KPMG’s view, how significant is this shift?

VK: It helps position SAP differently in the market. Before, the north star was SAP Cloud ERP. Now it’s the autonomous enterprise. That moves SAP away from being perceived solely as a system of record and toward becoming a system of contextualization, reasoning and execution.

SAP doesn’t only hold the business context. SAP is also saying it can provide the technical solution—meaning AI agents execute within the SAP environment rather than relying solely on ecosystem partners to provide those agents.

MG: Many enterprises are undergoing some form of AI transformation. The growth in capabilities is exponential. People are already experiencing this in their private lives, and those experiences are shaping expectations inside companies.

Then they return to the enterprise and encounter structured user interfaces [UIs] that require significant enablement and are often rigid. While enterprise systems can provide end-to-end process optimization and orchestration when implemented well, the experience isn’t always intuitive, especially when a process is only partially covered within the system and partly executed outside it.

Q: How does KPMG define the autonomous enterprise? Where is the line between AI-assisted and AI-executed work?

MG: With the word ‘autonomous,’ the intent is for processes and applications to become more self-executing, self-monitoring, self-observing, and self-correcting. At a minimum, it means achieving the highest degree of responsible agentic AI execution.

The question is how much automation, reasoning, and human-like work can be performed by agentic AI in a way that is responsible, trusted, reliable, and auditable.

Autonomous doesn’t mean handing over the keys to agentic AI and conducting spot checks later. It means gradually increasing the amount of agentic AI execution in a responsible way.

VK: We don’t define the autonomous enterprise as a fully self-running system. Autonomous doesn’t mean uncontrolled.

AI is increasingly augmenting, orchestrating and executing work, but humans remain accountable. SAP has launched toolsets that technically support this, including the AI Agent Hub, agent mining in SAP Signavio and the broader SAP Business Transformation Management [BTM] suite. However, organizations should still define policies and embed them into their operating model.

We also distinguish between maturity stages. Some work is fully human-executed. Some is AI-assisted. Some is AI-augmented. At the end of that spectrum, some work becomes AI-executed, where an agent performs a task with a higher degree of autonomy.

Q: KPMG US research has cited agentic system complexity as a barrier to scaling. What readiness gaps do you see with SAP customers?

MG: Enterprises often have an AI strategy, but in reality, they may have multiple strategies that aren’t connected. There may be a CEO-level mandate to implement AI across the enterprise, or a set of initiatives in various business functions. There may also be an IT and technology-driven AI strategy. That’s where the gap can appear. Companies should adopt a holistic view that connects the strategic intent, the technology platform, the people transformation, and governance processes. If those pieces aren’t aligned, organizations can face the risk of sub-optimizing parts of the transformation and losing trust in the outcome.

VK: From a people-centric point of view, we typically focus on five areas with customers. First, we help them reimagine work, not just individual roles. The question is what work exists within the organization, how it gets executed and by whom. Second, we help them invest in workforce readiness. That includes enablement, reskilling, upskilling, and helping employees become more comfortable with AI. Third, we help create new governance roles, accountabilities and governance models. Fourth, we help redefine career paths, particularly as traditional entry-level work is increasingly affected. And fifth, we support proactive change management.

MG: Another important consideration is day two. Everybody focuses on the lead-up to day one, meaning the go-live of an agent or AI capability. But are you ready for what comes next? If an agent is live, who monitors it? Who improves it? Who decides whether it’s producing the right business outcome? Who changes the process if the agent reveals a better way of working? That operating model should be ready as well.

Q: How should organizations think about governance before they start managing agents at scale?

VK: Many customers address governance too late. They implement something and only then begin discussing governance. Our view is that governance must be a foundational consideration from the beginning, especially if an organization wants to scale.

Data and process governance become even more important depending on the sector and region. If a company operates in defense, the public sector, critical infrastructure, or other regulated industries, or in regions with specific sovereignty requirements, governance isn’t optional.

Q: What should SAPinsiders be doing now to be ready for where SAP is heading?

MG: First, gain clarity on the roadmap and the existing capabilities. Keep your innovation radar open. Regularly review what’s new, what’s changed, what’s available in the current release, and which capabilities can now be activated that were not possible before.

Second, actively manage the innovation funnel. Review established processes and existing implementations. Survey functional leaders and end users to identify capability gaps and opportunities.

Third, institutionalize the AI conversation. Make it a regular part of how IT and the business engage with one another. Organizations should have a way to continually reassess what’s possible and connect emerging capabilities to business priorities.

Q: What would you caution SAP customers against?

MG: Be careful not to chase shiny objects and assume agentic AI is the answer to everything.

Likewise, don’t treat a broken process as an AI use case when it can be addressed through traditional process improvement. If data is entered into a system, exported to a data sheet, routed through email, and managed as an exception, the first step should be using process mining and process intelligence to find where the process is breaking down.

Agentic AI should be used where it materially improves the user experience, simplifies work, reduces the need to learn new interfaces, or allows people to interact with a process in a more natural way. Organizations shouldn’t use agents to mask a fundamental process issue that should be fixed within the underlying system.

VK: Be careful not to reduce the autonomous enterprise to a technology discussion. SAP has provided a clearer technology direction, but customers should still define their maturity journey. Which work should remain human-executed? Which activities should be AI-assisted or AI-augmented? Which processes are ready for AI execution? Those are business decisions, not just technology ones.

The organizations that are likely to make the most progress are those that connect strategy, data, process, governance, and people from the start.

Some or all the services described herein may not be permissible for KPMG audit clients and their affiliates or related entities.