For most of the last two decades, "digital transformation" had a fairly consistent meaning: replace paper and spreadsheets with a proper system of record, usually an ERP, and redesign processes around it. That work isn't finished at most companies — but the goalposts have started moving again, and generative AI is the reason why.
From systems of record to systems of judgment
A traditional ERP implementation answers the question "where does this data live, and how does it flow through the business?" It's exceptional at recording what happened. What it was never designed to do is exercise judgment — deciding which of a thousand open invoices needs attention today, drafting a first response to a customer complaint, or summarizing a claim file for an adjuster in thirty seconds instead of thirty minutes.
That's the gap generative AI is stepping into. Not by replacing the system of record, but by sitting on top of it, reading the data already captured there, and doing something a traditional ERP workflow couldn't: producing a judgment, a draft, or a recommendation in natural language, grounded in your actual business data.
What this means for a transformation roadmap
If your digital transformation roadmap still ends at "successfully implemented S/4HANA," it's worth extending it. The organizations getting the most value right now aren't treating generative AI as a separate initiative bolted on afterward — they're planning the AI layer alongside the ERP layer from the start, which is a large part of why we build SAP BTP and SAP Business AI into implementation conversations from day one rather than as a phase two.
Practical starting points we see working
- Copilots for internal workflows — assistants that can answer "what's the status of this order" or draft a summary of an account, grounded in live SAP data.
- Document and unstructured data processing — extracting structured data from contracts, invoices, or claims documents that previously required manual entry.
- Agentic workflow automation — multi-step processes that used to require a person to move data between systems, now handled end-to-end with human review at key checkpoints.
- Change management, not just technology — the biggest transformation risk isn't the AI model, it's whether the organization trusts and adopts it. That's a change management problem as much as a technical one.
The honest caveat
None of this works without the unglamorous foundation underneath it — clean, integrated data, sensible governance over what an AI system is allowed to act on, and a Clean Core-aligned architecture that keeps the whole thing maintainable. Generative AI amplifies whatever foundation you already have. If that foundation is fragmented, the AI layer will surface that fragmentation faster than anything else would.
The organizations doing this well aren't chasing every new AI capability as it appears. They're picking a small number of high-value use cases, grounding them properly in real business data, and building outward from there — the same disciplined approach that's always separated successful digital transformation from expensive experimentation.