Javier Rubio. CPO at comerzzia
I was coming back from a customer meeting with a company facing a great deal of change, with a certain thought still going round in my head since the weekend, after reading an article by SAP CEO Christian Klein entitled “The AI Race Is Being Fought in the Wrong Place”. While I fully agree with the role of ERP and its processes as key enablers of efficient AI within companies, reading between the lines left me thinking about the moment we are living through and the industry narratives surrounding the adoption of artificial intelligence.
What caught my attention was the suggestion of a kind of “tension” between early AI adoption and ERP. In the background, there seems to be an underlying idea that renewing or migrating the ERP is essential as the vehicle for achieving a productive and orderly use of AI. On this point, I would like to share some of the questions and reflections I am increasingly coming across with my customers.
1. The commercial narrative vs. the technical reality
ERP should be the backbone. That much is true. But without the need for migration, the current SAP versions used by most of my customers already contain the processes, operations management and structured data model that, in many cases, also allow the use of APIs or REST services. Broadly speaking, this is what we bought an ERP for.
Technically, an existing ERP can already act as the “source of truth” for specialised AI agents, provided that its data and actions are exposed through APIs. Migrating to a new version should not be a mandatory requirement in order to start extracting value from AI.
2. The “hijacking” of innovation and the cost of waiting
I have a significant number of customers currently involved in SAP upgrade processes, with estimated timelines of between 18 and 24 months, who have decided to put the rest of their technology projects on hold. This raises a critical dilemma: should we pause AI until the migration is complete?
Waiting two years creates a major operational risk. AI adoption is not just a technical matter; it is also a change management process. Those who start in 2028 may have a very modern core system, but they will have lost two years of cultural and operational maturity compared with competitors who did not stop.
3. Domain-based modernisation and hybrid ecosystems
Do we really need to tackle everything at once? I believe it can be far more efficient to renew by domains or business areas while, in parallel, starting to make use of agents, APIs and protocols such as MCP, the Model Context Protocol.
The reality is that most companies do not operate within a single-vendor ecosystem. Multiple specialised best-of-breed solutions coexist and connect with the ERP. An intelligent and decoupled AI layer would allow us to connect those worlds today, without having to wait until the end of a large monolithic migration.
4. Embedded AI: does it give me wings or hold me back?
Finally, I find myself considering a strategic question: is relying exclusively on the embedded AI offered by the ERP vendor an advantage or a limitation?
On the one hand, it gives us “wings” by offering immediate, natively connected solutions to standardise basic processes. But on the other, it can also limit us if it ties us to the vendor’s innovation roadmap and makes us look like the rest of the market. If we all use the same standard AI to make decisions, where does our competitive advantage remain?
Perhaps the answer is not to choose between AI and ERP, but to adopt a hybrid model: use the standard for the basics, while building our own intelligence in the areas that make us unique. And, of course, migrate to a new version where appropriate, with the right re-engineering and redefinition of processes.
What do you think? Are we delaying AI adoption because of major corporate migration projects?
Javier Rubio
CPO at comerzzia
