AI in Shared Services: why automating the existing process is the wrong shortcut

The wave that is washing away what has been built

To understand why AI calls for a redesign, it’s worth looking at what is happening simultaneously in large-scale operations.

Many companies that have invested heavily in hyperautomation over the past ten years, with fleets of 200 or 300 RPA (Robotic Process Automation) robots, are now in the midst of ERP transitions. To SAP’s S/4HANA in many cases, or even to other equivalent platforms in others. In all of these cases, the migration changes the organization’s foundations and often disrupts the previous automation ecosystem.

Those going through this process quickly realize: recreating the same robots on the new platform can often be a waste. The question then becomes: what capabilities does the platform itself already deliver natively, where does AI add real value, and where does traditional automation still make sense?

This is where AI comes in, and in a different way than one might expect. Instead of being just another layer to accelerate the existing process, AI reopens the discussion of mandate. Who decides what. How CSC, business areas, and IT share responsibility and deliver. How governance and security gain pillar status, after years of being treated as an appendix.

The most common mistake: applying AI to the old process

Here is the most sensitive part of the conversation, and what I heard most often both on the SSOW stages and at the CSC Summit panel.

When an organization applies AI to an existing process, it tends to end up with a slightly faster version of that same process. It gains minutes, perhaps hours, but misses out on the real opportunity that the technology offers: rethinking the design based on what AI makes possible.

A practical example from our transformation work. Traditional process mapping takes weeks of interviews, workshops, and flowcharting. Today, with well-structured context databases and prompt engineering, we’re able to arrive at the first project meeting with a starting process already mapped out, based on knowledge the organization already possesses but which was previously scattered. Beyond mapping, we have the process of analyzing processes, ideation and refinement of solutions through multiple lenses, specification and design, and even the implementation itself. Operational work shrinks; analytical work gains ground.

The same logic applies to change management. In the past, simulating the organizational impact of a change required weeks of spreadsheets and meetings, as well as a limited view of the organizational dimensions; however, today, starting from a transformation hypothesis, it is possible to automatically map impacts on processes, structure and functions, systems, regulations, policies, results, operational KPIs, customer metrics, and more.

The benefit here goes beyond speed, because what changes is the lens. We’ve moved away from the effort to better answer the questions we already know how to ask and have started to formulate questions we didn’t know or that previously didn’t fit within the budget or timeline of initiatives.

What sets apart those who are ready to take the next step

Let me return to the image that kicked off this discussion. AI is the new lens, but the lens is worthless without what it needs to see.

For Brazilian CSC/GBS teams, this means three key areas that cannot be delegated to IT:

• Integrated Database:a defined service catalog, and reliable data and master records. Without this, AI inherits the noise.

• Clear mandate: the structure must be at the table where company strategy is decided; it should not be called in later just to “implement.” Deep business knowledge, beyond transactional processes, is non-negotiable.

• Renewed ways of working:leveraging AI to support the analytical discipline we already master, without letting it replace root-cause investigation. What changes is the lever for execution, not the way of thinking.

   

The question for 2026 is simple to ask but difficult to answer: Which processes in your CSC or GBS would you design today, from scratch, if AI were already part of the starting point? How much of what currently exists would survive this question?

The GBSs that manage to answer this question without shielding themselves end up shifting from the role of strategy executors to that of strategy designers, and this shift may be the best definition of maturity in shared services today.