"Our Joule assistants and agents are designed for cloud. This is destiny and destiny is unchanged."
— Sebastian Steinhaeuser, Chief Strategy Officer, SAP
At Sapphire 2026, SAP reversed a long-standing stance and announced that Joule AI agents would finally be available on ECC and on-premises S/4HANA systems, not just in the cloud. If you are one of the more than 20,000 SAP customers still running ECC, this sounds like relief: deploy the AI, keep the system you already know, and quietly let the migration slide. It is not relief. The deadline has not moved, the AI offer comes with a cloud-commitment string attached, and treating agents as an alternative to migrating is exactly the mistake that will leave you stranded. Here is what actually changed, what did not, and what to do about it.
What SAP Actually Announced
SAP's new offer lets Cloud ERP Private Edition customers activate up to three Joule Assistants at no additional cost, covering finance, HR, and supply chain workflows. The extension to ECC and on-premises S/4HANA customers looked, on the surface, like SAP finally meeting legacy customers where they are. It is not that simple. Access to Joule on ECC is only granted if the customer commits at least 50% of their maintenance spending to the cloud first, alongside signing up for SAP's Max Success Plan. SAP's Chief Strategy Officer Sebastian Steinhaeuser was explicit about why: "our Joule assistants and agents are designed for cloud. This is destiny and destiny is unchanged."
In other words, the AI is not a substitute for the migration, it is a lever to accelerate it. SAP's own Chief Product Officer for Business Suite, Manoj Swaminathan, framed the intent plainly: "we want to take every customer along with us." That is a company trying to pull its legacy base toward the cloud using AI as the incentive, not a company offering ECC customers a permanent AI-powered off-ramp from migrating at all.
The Core Misreading to Avoid
"SAP now supports AI agents on ECC" does not mean "ECC is now a viable long-term AI platform." It means SAP built a bridge with a toll on it, priced in cloud commitment, to get you off ECC faster. Read the announcement as a migration accelerant, not a migration alternative, and the rest of your planning gets much clearer.
The Deadline That Did Not Move
While the AI headlines were dominating Sapphire coverage, the fact that mattered most was quietly unchanged: SAP ECC mainstream maintenance still ends December 31, 2027, with extended maintenance available only through 2030 at a roughly 9% premium on standard maintenance fees, translating into millions of dollars annually for large enterprises. No new features arrive on that extended-maintenance track, and critically, no access to Joule agents, SAP Business Data Cloud, or the broader AI Platform is included in it. We covered the deadline mechanics in full in our breakdown of the ECC end-of-support timeline, and nothing in the Sapphire 2026 AI announcements changes a single date in that piece.
The math around that date is unforgiving. A full S/4HANA migration runs 18 to 36 months depending on approach, greenfield implementations at the fast end around 12 to 18 months, brownfield conversions at 18 to 30, and selective data transition stretching to 24 to 36. One industry analysis put it bluntly: "if you have not started your migration by mid-2026, you are unlikely to complete it cleanly before the December 2027 deadline." We are past that line already. Every quarter of delay from here compounds against a fixed date that is not going to slip, a dynamic we detailed in our guide to choosing a migration path and building AI readiness.
| Fact | What It Means |
|---|---|
| ECC mainstream support ends | December 31, 2027, unchanged by any AI announcement |
| Extended maintenance | Through 2030, ~9% fee premium, no new features, no Joule/AI Platform access |
| Joule on ECC condition | Requires committing 50% of maintenance spend to cloud, plus Max Success Plan |
| Customers still on ECC | 20,000+ per one industry estimate; up to 40% may remain by 2030 per another |
| Full migration timeline | 18-36 months depending on approach; started mid-2026 at the latest for a clean 2027 finish |
Why "Just Add Agents" Does Not Solve the Underlying Problem
There is a deeper reason AI agents cannot substitute for migration, and it is architectural, not contractual. Agents, whether SAP's own Joule or a custom-built assistant, are only as reliable as the data and interfaces beneath them. An ECC system has none of the clean, governed foundation that safe agent autonomy actually requires. We made this case in detail in how to deliver a clean core in S/4HANA: clean, stable APIs and disciplined data governance are what let an agent act correctly, and ECC's decades of accumulated customization are precisely what a clean-core architecture is built to eliminate. Bolting Joule onto ECC does not retroactively clean the data it has to reason over.
This is also why third-party vendors are stepping into the gap SAP's condition creates. MyWave, led by former SAP president Geraldine McBride, now offers AI agents running natively on ECC without requiring any cloud commitment at all, positioning itself directly against SAP's 50%-cloud-spend condition. That competitive pressure is real, and analysts have noticed it. IDC's Mickey North Rizza called AI "the new carrot and stick" reshaping how vendors move customers through their lifecycle, and Lopez Research's Maribel Lopez was candid about the tension: SAP is trying to offer cloud-designed AI on-premises, and "SAP needs to find a way to migrate customers faster." The agents are a negotiating chip in a much larger vendor-versus-customer standoff over migration timing, not a technical solution to the deadline.
The Uncomfortable Number
One analyst estimate puts the share of SAP customers still on ECC by 2030 at roughly 40%, with, in that analyst's words, "a good third" not even actively thinking about migration yet. If that holds, a very large share of the SAP installed base is planning to run on emergency extended maintenance, without new AI capabilities, well past the point where competitors who migrated on time are running agents against a governed, current system. Whatever AI agents ECC eventually gets access to under SAP's cloud-commitment terms, the customers who never migrate at all get none of it.
Where AI Agents Genuinely Help: Accelerating the Migration Itself
None of this means AI agents are irrelevant to your 2027 problem, they are simply pointed at the wrong target if you deploy them to prop up ECC instead of to get off it faster. The more useful application is agent-led migration tooling aimed squarely at the migration project, which SAP itself is now promoting alongside third-party toolchains: custom code analysis and ABAP-to-BTP conversion recommendations, automated data quality assessment, and cleansing orchestration, the exact categories of work that historically eat the most calendar time in a brownfield conversion. We covered the mechanics of this in how AI is cutting S/4HANA migration timelines from years to months, and the reported effort reduction from agent-led tooling, upwards of 35%, is large enough to meaningfully compress an 18-to-36-month timeline that most organizations no longer have slack in.
Some vendors are running dedicated agent toolchains across the entire SAP Activate methodology, from discovery through cutover, and reporting that a focused pilot can go from kickoff to measurable hours saved on a single transaction within 90 days. That is the right instinct: point the agent at removing manual migration work that is competing for the same scarce specialist capacity everyone else is also bidding for. Talent scarcity is not a minor factor here either; roles like Universal Journal-certified finance consultants, BTP extension developers, and experienced cutover leads are already in short supply, and day rates are rising as more organizations start their projects late in the same shrinking window.
The Decision Framework
Strip away the marketing and the choice in front of any ECC customer is actually simple to state, even if it is hard to execute. If you have already started migrating, the right move is to point every AI agent budget line at compressing the remaining timeline: custom code remediation, data cleansing, and testing automation, not at building a nicer ECC. If you have not started, the AI news from Sapphire is not a reason to wait longer, it is a reminder that SAP itself is now actively pricing you toward the cloud, and every quarter you delay narrows your realistic path to a clean, on-time cutover before December 2027. And if you are seriously considering staying on ECC past 2027 via extended maintenance, go in with eyes open: no Joule, no Business Data Cloud, no AI Platform, a 9% fee premium, and a hard stop in 2030 that will eventually force the exact same decision under worse conditions.
The honest reading of SAP's Sapphire 2026 AI-on-ECC announcement is that it changes the incentive structure around migration, not the deadline itself. Treat the agents as what SAP built them to be: a way to migrate faster and with less manual pain, not a way to avoid migrating at all. The wall in front of every ECC customer is still December 31, 2027. No AI agent, from SAP or anyone else, moves it.