The Capability Gap Is Real — And It’s Not a Reason to Wait
The standard ERP transformation story has a fixed order: migrate first, then unlock AI. That order is no longer defensible, and the cost of following it is compounding for every quarter you do.
The real problem isn’t that organisations have to migrate eventually — most do, and the case for it is only getting stronger as SAP maintenance costs escalate outside RISE with no ceiling. The problem is the assumption that AI value has to wait until the migration is finished. It doesn’t, and treating it that way is what makes most transformation business cases weak.
Competitors who’ve already moved to S/4HANA are running Business AI today: automated close, demand forecasting, and intelligent procurement. Every quarter spent on the old platform without an AI plan, that gap widens. That’s the cost case for moving.
But there’s a second, quieter cost that gets missed: most transformation programmes don’t bring AI into the room until the platform is already built. By the time the new system goes live, the data structures and integration points are fixed, and AI gets retrofitted into a design that was never built with it in mind. Retrofitting costs more and delivers less than building it in from day one.
That’s not an argument against migrating. It’s an argument against deferring AI to a phase that comes after.
The Information You Need Is Already in Your System
There’s an industry assumption that enterprises have an AI context problem — that the data and processes aren’t well enough understood for AI to be useful yet, and the fix is another platform layer on top.
That’s not where the real bottleneck is. The business context is already sitting inside the ECC environment: in the configuration, the master data, the workarounds nobody documented but the team still runs on. A programme that brings AI in at the start can read that context directly, use it to scope real use cases, and start activating some of them before go-live — not after.
That’s the difference between AI as a phase-two add-on and AI as something that’s quantified, scoped, and already producing value while the rest of the programme is still in flight.
Why This Changes the Business Case
Cost avoidance alone rarely produces a business case a CFO finds compelling. Migration costs land in year one; the savings from coming off ECC maintenance take years to catch up, and the payback period often stretches to year three or four. That’s the conversation where someone reasonably asks why you’re spending millions to save a few hundred thousand a year.
The case gets stronger the moment AI use case value is counted alongside cost avoidance — not as a future-state promise, but as value that starts accruing during the programme itself, before go-live. That’s a different conversation with the board: not “what does it cost to stop the bleeding,” but “what does the platform let us do that we can’t do today, and how much of that can we capture while we’re already building it.”
What This Means in Practice
Three things follow from this, and none of them are about skipping migration:
Bring AI into scope from day one, not as a separate workstream that starts after cutover. The people who understand the AI use cases need to be in the room while the data structures are being designed, not retrofitting around decisions that are already locked.
Quantify the AI case before committing to the programme, not after. A credible business case names specific use cases, by department, with a realistic timeline for when each one starts paying back — not a generic “AI will help” line item.
Treat clean core and AI activation as the same discipline, not sequential ones. A programme that enforces clean core throughout — not as an aspiration retrofitted at the end — is the same programme that can activate AI cleanly, because the same governance that keeps custom code under control is what keeps AI integration points stable.
The organisations that get the most out of their next ERP investment won’t be the ones that migrated fastest. They’ll be the ones that stopped treating AI as something that starts after go-live.
