Table of Contents
For years, the enterprise software industry used one word to describe what ERP systems do: record. They record what was purchased, what was produced, what was shipped, what was posted. The data lives there. The truth lives there. But the work, the decisions, the fixes, the adaptations, happens somewhere else. In spreadsheets. In inboxes. In meetings.
At Sapphire 2026, SAP’s Christian Klein made something explicit that many of us in the SAP ecosystem had been sensing for some time. The next chapter for ERP is not about storing the truth, It’s about acting on it, faster, smarter, and increasingly without waiting for a developer to build something first.
He called it the Autonomous Enterprise. We think it’s the most important strategic bet SAP has made in a decade, and also the one most likely to be misunderstood.
Enterprise AI is not a chatbot problem. It is an operational execution problem. And those are very different things to solve.
The Framing That Cut Through
“Systems of record” to “systems of action”. This distinction landed harder than any product announcement at Sapphire. Not because it’s technically new, but because it’s honest in a way that most enterprise software marketing isn’t. It admits that AI features bolted onto reporting layers don’t actually change how a business runs.
A system of action is different. It means AI that can trigger workflows, enforce approvals, adjust operations, surface exceptions, and more inside the business processes that already govern the company, not layered on top of them from outside.
The reaction from enterprise SAP leaders was immediate. Executives across industries debated what it actually takes to get there. A conversation worth having.
The Prerequisite Nobody Wants to Talk About: Data You Can Trust
“Most mid-market finance teams are still spending 40–60% of close time on manual reconciliation before the data is trustworthy enough to act on. Until that layer is solved, the autonomous enterprise remains aspirational.” — Maxim Munvez, CFO, manufacturing.
This is the prerequisite problem. Autonomous execution is only as reliable as the data it runs on. And in most SAP environments we’ve seen, especially mid-market and manufacturing, the data layer is not clean. It is semi-structured, partially reconciled, and maintained through tribal knowledge that lives in Excel files and the heads of long-tenure finance analysts.
Supply chain leaders echoed the same thing:
“All automations — AI, machine learning, optimizers look great on paper, but AI requires pristine data and end-to-end integrations. A penny saved is a penny earned doesn’t really apply here.” — Emil Juszczak, Supply Chain Process Manager
This is not a software problem, it is an organizational one. The good news: it is solvable inside SAP, without a multi-year data warehouse project, if the tooling allows business users to define their own reconciliation logic and apply it continuously, on live data, without waiting for IT.
The Governance Trap: Why Bolted-On AI Creates More Work
The second real obstacle to the autonomous enterprise is governance and specifically, the failure mode of AI that operates outside SAP’s existing authorization and approval structures.
Every SAP environment has years of governance logic built into it: segregation of duties, approval hierarchies, role-based authorizations, audit trails. When AI operates as an external layer, pulling data out of SAP, processing it elsewhere, pushing results back in, it creates overhead that compounds with every integration point.
“When AI is truly inside the ERP, adoption accelerates quickly. When it is bolted on from outside, you spend more time managing the seams than getting the work done. And if it is truly inside, I can see us eliminating a few audits altogether.” — Eric Sandberg, Enterprise Applications & ERP Transformation Executive
The distinction that separates realistic AI adoption from the wishful-thinking kind is AI that respects SAP governance by operating inside it is a trusted system. An AI that bypasses SAP governance by operating around it is a liability.
The regulated-industries perspective is even sharper:
“‘Inside the trusted system’ does not automatically mean ‘ready for autonomous execution.’ In regulated environments, the real test is whether AI-driven action can be validated, governed, supervised, and audited within the operating model.” — Claudia D., Founder, Execution Architecture for Boards & Owners
Auditability is a precondition for adoption in enterprises that take compliance seriously.
What “System of Action” Actually Looks Like Under Pressure
The abstract version of this argument is clean. The real version is messier and more instructive.
In mid-2024, Lumenis, a global medical technology company, received a mandate with a hard deadline: Boston Scientific had acquired the surgical division for $1 billion, and the deal required a full IT separation, one SAP environment for Boston Scientific’s surgical business, one for the new Lumenis aesthetic and ophthalmic divisions. Three and a half months. No extensions.
This is the kind of operational event where the gap between “system of record” and “system of action” stops being a strategic concept and becomes a concrete problem measured in hours.
Every module had to move: Material Ledger, inventory, accounts receivable, accounts payable, assets, service contracts, purchasing orders. Data had to be split with surgical precision, literally, between two companies, with controllers marking every line as surgical or non-surgical, and automated logic handling the rest. And it had to be done on live production data, inside SAP, with a monitor tracking what moved, what didn’t, and what needed to be reversed if something went wrong.
“We could not have done it without SAP and Insight.”
— Tahel Blum, Head of IT, Lumenis
The number that makes this concrete: $100 million in inventory moved in approximately three hours using automated background processing. Financial data (AP, AR, GL) took a few hours more.
What made that speed possible wasn’t a data lake, an external AI layer, or a custom ABAP development sprint. It was tooling that let experienced business consultants build the logic directly inside SAP, deploy a real-time monitor to track progress, and build a reverse function that could undo postings with a single click if something went wrong.
The governance was built into the process from the start. Every posting traceable. Every action reversible. Every controller accountable for their line.
That’s what a system of action looks like when it works.
Where DPRO Sits in This Conversation
We built Insight Zap and Zappy AI because we watched this problem from inside SAP customer environments for years. The pattern was always the same: the data was in SAP, the governance was in SAP, the business logic was in SAP, but the actual work was happening in spreadsheets and workarounds outside it. Every AI layer added externally made the drift worse, not better.
What Zappy AI actually does:
Zappy AI is not an AI assistant that sits next to SAP. It is an operational layer that sits inside SAP, on live production data, under existing SAP authorizations and governance. Business users in finance, supply chain, and operations can ask questions, build reports, configure workflows, and define their own reconciliation logic without writing ABAP, without opening a ticket, and without waiting for a developer to interpret what they need.
The reconciliation logic that a finance analyst has been maintaining in Excel for six years can be moved inside SAP and automated — on the fly, without rebuilding underlying models. That is what solving the data readiness layer looks like in practice.
The governance piece was non-negotiable from day one. Zappy AI operates within existing SAP authorizations and approval structures — not around them. There is no shadow authorization layer. There is no data leaving the system.
What the Next Generation of Enterprise AI Will Actually Be Judged On
The conversation at Sapphire — and everything that followed it across the SAP practitioner community — points to a simple conclusion: the market is shifting its evaluation criteria for enterprise AI.
The old criteria: how many use cases does it support? How good are the answers? How fast can you implement?
The new criteria:
- Does it operate on data you can actually trust? Or does it inherit the reconciliation backlog your team has been managing manually for years?
- Does it work within your governance, or around it? Can you put it in front of an auditor and explain exactly what it did and why?
- Can business users operate it directly? Or does every change require a developer, a ticket, and a three-sprint delay?
- Does it reduce the distance between SAP and where work actually happens? Or does it add another disconnected layer to manage?
The autonomous enterprise is not coming all at once. It will arrive company by company, in the organizations that solve these four questions first, not in the ones that buy the most agents.
The next generation of enterprise AI won’t be measured by how many agents exist. It will be measured by how safely and quickly businesses can adapt operations inside the systems they already trust.
FAQ
At Sapphire 2026, SAP CEO Christian Klein articulated the Autonomous Enterprise vision, a shift from ERP as a system of record to ERP as a system of action. This means AI that can execute decisions, trigger workflows, and adapt operations in real time, inside existing SAP governance and authorization structures, not as a disconnected AI layer operating outside them.
The most consistently cited obstacle is data readiness. AI cannot execute reliably on data that hasn’t been reconciled and validated. In most mid-market SAP environments, significant time is still spent preparing data manually before it can be trusted for automated action. Solving this prerequisite layer, inside SAP, is the first real requirement for autonomous execution
AI inside SAP operates within existing authorizations, governance rules, audit trails, and approval structures. It acts on live production data with full traceability. AI built around SAP creates additional governance risk, latency, and “seam management” (the overhead of keeping two systems synchronized) rather than acting within one trusted system
Insight Zap is DPRO’s in-SAP operational layer. It allows business users to build reports, configure workflows, and act on live SAP data without ABAP development.
Zappy AI, its AI component, enables natural language queries, automated reconciliation logic, and configurable workflows, operating under existing SAP authorizations and governance. It addresses both the data readiness prerequisite and the governance requirement for autonomous enterprise.
