Can AI Agents Replace SAP? The Enterprise Software Cage Match We Didn’t Know We Needed

At the AI Infrastructure Summit, I threw a curveball at Luke Norris crew. The kind of spicy question you ask not just to provoke debate, but to surface the fault lines in our collective assumptions: Can a non-deterministic AI system replace a system of record like SAP?

Luke, cool under pressure, offered he calls the “Run It Back Turbo” defense. When the answer isn’t clear, the AI agent keeps rerunning until it’s confident. And when determinism is required—say for accounting or supply chain calculations—the agent writes the code on the fly. Even the verification logic. In theory, it creates and executes deterministic code from within its non-deterministic core.

Sounds slick. But enterprise IT folks know better than to be seduced by theory alone.

SAP Is More Than Code

Enterprise Resource Planning (ERP) systems like SAP are not simply deterministic machines of business logic. They are deeply integrated reflections of an enterprise’s culture, compliance requirements, operational assumptions, and—let’s be honest—bureaucratic inertia. They are the system of record.

We’re not just talking about running payroll. We’re talking about reconciliation processes required by regulators. We’re talking about supply chain coordination that spans continents. We’re talking about embedded risk models that determine whether your CFO signs off on quarterly earnings or faces a fine. These are not things you want a GPT-5 agent “figuring out” in the moment.

And yet… this AI agent model is tantalizing. Imagine an AI that, when presented with a new regulatory requirement, writes the workflow logic, implements data validation rules, and spins up compliance dashboards. All without waiting for the next SAP enhancement pack.

That’s the dream.

AI Agent vs. Intelligent ERP: Who You Got?

SAP (and its competitors) have spent the past decade pitching “intelligent ERP.” That term is code for: “We finally made our systems usable enough that the business can ask hard questions and expect real answers” . Think predictive analytics, in-memory processing, and embedded ML algorithms.

But intelligent ERP still requires deterministic thinking. The system answers questions like “Who needs our help the most?” only after a massive lift: integrating systems, aligning schemas, securing data flows, and training business users on query tools.

AI agents flip the script. They assume the chaos. They thrive in ambiguity. They pull data from wherever it lives, figure out the question you’re trying to ask, write the integration glue, and—most provocatively—assert an answer with confidence.

It’s not quite hallucinating. But it’s also not the ISO-certified, auditor-approved workflow that SAP spent decades perfecting.

So… Can AI Replace SAP?

Technically? Maybe.

Practically? Not yet.

Culturally? Absolutely not.

We trust deterministic systems because we understand them—or at least we know someone who does. The entire audit profession, for instance, is built on the assumption that every financial transaction has a provenance. AI agents, with their probabilistic reasoning, aren’t just breaking the model. They’re rewriting it.

Luke Norris is right in one sense: when math is required, the agent will write and verify the math. But who signs off? Who owns the code that just wrote itself? And what happens when it changes tomorrow?

If we go down this road—and I suspect we will—it won’t be a replacement of SAP. It’ll be the beginning of a new class of systems: Probabilistic Systems of Intelligence. They won’t replace our systems of record. They’ll sit above them. Translate them. Interpret them. Challenge them.

But replace them? SAP can sleep easy. For now.

What do you think? Would you trust an AI agent to run your quarterly close? Or are we still a generation away from letting non-deterministic systems run our deterministic world?

#AI #ERP #SAP #IntelligentERP #EnterpriseIT