Table of Contents
Key Takeaway
Before answering whether your SAP environment is ready for AI, answer five operational questions: where your data actually lives, how long adaptation cycles take, what’s sitting in your backlog, what AI access you have today without a migration commitment, and what an agent would actually find if deployed tomorrow. If three or more of the five reveal a gap, the readiness conversation needs to happen before the AI conversation — not after.
Most SAP leaders now have the same item on their agenda: the AI story. The board saw the Sapphire announcement. 224 agents. 51 Joule Assistants. Autonomous Enterprise. Now the question is on the table. Before you answer it, answer these five first. They are not technical questions. They are operational ones. The answers will tell you more about your actual AI readiness than any vendor demo.
Question 1: Where does your operational data actually live today?
Most SAP environments have operational data living outside SAP itself, in spreadsheets, BI exports, and email threads. An agent can only act on data it can access, so any decision-critical data outside SAP is invisible to it.
Pick the five decisions your team supports most often. For each one, where does the data actually come from? Not where it is supposed to come from. Where it actually comes from. Write it down.
In most environments, the honest answer has more than one source. SAP is one of them. The Excel file finance maintains for month-end is another. The BI export the operations team runs before the morning standup is another. The email chain where procurement tracks supplier exceptions is another.
This is the first question because everything else follows from it.
Question 2: How long does it take to respond to a new operational requirement?
The time it takes your SAP team to deliver a simple operational change, a new column, a new filter, a calculation that doesn’t exist yet — is the clearest predictor of how many workarounds already exist outside SAP.
Not a major development. Something simple. Write down a realistic number. Days? Weeks?
That number is the speed gap. It’s also the origin story of your workarounds: every time the answer was weeks, the business made a choice, wait, or build something outside SAP. Many chose to build. That something became permanent, and over time it became the operating model.
From what we see at DPRO: when the adaptation cycle exceeds two weeks consistently, operational teams start building permanent workarounds within months. Within a few years, those workarounds are the real operating model, and SAP is the system the workarounds report to. Agents cannot reach the workarounds. They can only reach SAP.
Question 3: What is actually on your backlog right now?
A SAP backlog full of operational adaptation requests, not bugs, not migration work, but requests to change a report or calculation, is a direct signal of how much of the business is already working around SAP.
Go through your top 20 open items. Sort them into three groups: system issues and bugs, operational adaptation requests (new reports, updated calculations, changed views), and architecture and migration work.
The second group is the one that matters most. Each item in it represents a place where the business is currently working around SAP because it cannot wait for the queue to clear.
If your backlog has “update FBL1N logic” (a standard SAP transaction for vendor line-item display) sitting in it for two years, that is not a development item. It’s a signal that finance is working around a report SAP already owns. Closing that gap doesn’t require new development, it requires giving finance a way to adapt that report themselves, inside SAP, without opening a ticket. Those items are the first places to close the gap before the agents arrive.
Question 4: What is your AI access today without a migration commitment?
SAP’s AI entitlement is tied to the cloud path. RISE with SAP (SAP’s subscription bundle for moving core ERP to the cloud) customers get a contractual commitment to three Joule Assistants activated in Year 1. ECC (SAP’s older on-premises ERP suite, the predecessor to S/4HANA) and on-premises S/4HANA customers get conditional access, available only upon commitment to move the majority of the landscape to SAP Cloud ERP.
This is the Sapphire detail most recaps didn’t explain clearly.
What’s actually available today, without that commitment: SAP’s Custom Code Migration Agent (part of its broader ABAP AI capabilities) reached general availability in Q2 2026. It analyzes custom ABAP code, identifies S/4HANA compatibility issues, and generates remediation recommendations. The underlying model was trained on more than 250 million lines of ABAP code, and SAP has stated a target of a 40% efficiency gain in custom code transformation work, a figure specific to this agent, distinct from the migration-effort numbers tied to SAP’s separate cloud migration partnerships. For ECC teams with heavy custom (Z-code) development, this is a meaningful tool, not for day-to-day operations, but for building a quantified migration business case.
Everything else, the Autonomous Close, the Demand Planning Agent, the full Joule suite, is on the cloud path. Zappy AI, by contrast, is built to work with SAP as it exists today, without requiring a migration commitment first.
This is not a reason to accelerate or delay a migration decision. It’s a fact to have clearly in hand before the steering committee asks what your AI roadmap looks like.
Question 5: If you deployed an agent tomorrow, what would it actually find?
An agent deployed today can only act correctly if the SAP authorization model behind it is current, not the roles configured at go-live and never revisited since.
Pick your highest-value operational use case. Walk through what data that decision requires. Is all of it currently inside SAP, current, and accessible? Or does it require pulling from multiple transactions, cross-referencing a spreadsheet, and a human who knows how to interpret the gaps?
One more thing most AI rollout plans don’t address: the authorization model. SAP’s agents operate inside existing SAP authorization objects, a Buyer-role agent cannot approve a purchase order beyond the buyer’s authorization limit. The governance model is already wired in, and that’s good news. But for it to work correctly, the authorization model has to be clean. In most ECC environments running a decade or more, roles accumulate drift: permissions configured at go-live, never revisited, users with access that no longer matches their actual function.
Before the agents work, the authorization model they run on has to reflect how the business operates today, not how it operated at go-live.
What the answers tell you
Two outcomes are possible.
The first: your operational data is mostly in SAP, your adaptation cycle is reasonable, your backlog is manageable, and your authorization model is reasonably current. You’re closer to agent-ready than you probably think. The gap is real but closeable in months, not years.
The second: the answers reveal a significant gap, multiple critical processes running outside SAP, a long adaptation cycle, a backlog full of workaround-prevention requests, authorization drift. As a rough threshold: if three or more of the five questions point to a gap, the AI conversation isn’t the urgent one, the readiness conversation is. That conversation is more valuable to have now, before the pressure to deploy intensifies.
Either way, you now have a real answer for the steering committee. A clear picture of where you actually stand.
That is the only place to start.
Curious where your own environment lands on these five questions? DPRO’s SAP Business Agility & Value Assessment walks through a version of this scoring in about ten minutes.
Related reading: SAP as a System of Action: What the Autonomous Enterprise Actually Requires · SAP Clean Core in Practice webinar
FAQ
Check five things: whether your operational data lives inside SAP or is scattered across spreadsheets and email, how long it takes your team to respond to a simple change request, what’s sitting in your backlog, what AI access you have without a migration commitment, and whether your authorization model is current. If three or more of these show a gap, focus on SAP readiness before the AI rollout.
RISE with SAP customers have a contractual commitment to activate three Joule Assistants in Year 1, with a path to SAP’s full Autonomous Suite roadmap. This access is tied to the cloud subscription itself, not a separate purchase.
ECC and on-premises S/4HANA customers get conditional access to most of SAP’s AI portfolio, available only upon commitment to move the majority of the landscape to SAP Cloud ERP. The one exception available today without that commitment is SAP’s Custom Code Migration Agent, which reached general availability in Q2 2026.
It analyzes custom ABAP code, flags S/4HANA compatibility issues, and generates remediation recommendations. It’s built on a model trained on more than 250 million lines of ABAP code, and SAP has stated a target of a 40% efficiency gain in custom code transformation work. It’s a migration-planning tool, not a day-to-day operational agent.
SAP’s agents operate inside existing SAP authorization objects, so an agent can’t act outside a user’s actual permissions — for example, a Buyer-role agent can’t approve a purchase order past that buyer’s limit. If the authorization model has drifted from how the business actually operates, the agent inherits that drift, so it needs to be current before agents are deployed.
