AI-first redesign for an OEM

By Elisabeth Goos, Sebastian Gundermann and Daniel Rohrhirsch

Large organizations have long pursued operational efficiency through incremental process improvement – adjusting existing workflows at the margins rather than questioning their fundamental design. As artificial intelligence (AI) capabilities mature, this approach is showing its limits. Organizations that layer new technology onto fragmented, legacy processes rarely achieve the step-change they are looking for. The more consequential question is how processes should be designed from the outset with AI as a first principle. This case study shows how Roland Berger helped a leading global automotive original equipment manufacturer (OEM) develop a scalable process redesign approach anchored in outcomes and an explicit human-versus-AI responsibility split.

The case for a zero-based process redesign

Our engagement with a leading global automotive OEM – with revenues exceeding EUR 200 billion and operations in more than 30 countries – shows why incremental optimization was no longer enough. The client's processes had been improved over time, but they were still shaped by legacy systems and organizational structures rather than by the outcomes they needed to achieve.

This created a practical problem. High numbers of manual steps and approval loops, reinforced by coordination dependencies, made processes difficult to automate or redesign at scale. These issues were not simply inefficiencies in the existing process – they showed that the process logic itself needed to be rethought.

What was required was a "zero-based" redesign mindset: rather than starting from the existing process map, we started from the desired outcome, then defined the capabilities required, the activities suitable for AI and the points where human judgment remained critical.

The Roland Berger five-step AI-first redesign approach

We structured the redesign around five steps, making the approach easier to repeat across business units while keeping it anchored in the process outcome rather than in the existing workflow:

#1 Define the outcome

The first step establishes what end-to-end success looks like. Clear outcome statements and critical success factors provide directional guidance throughout the redesign, replacing the conventional starting point – mapping what currently exists – with a deliberate focus on the business intent and what the process is ultimately supposed to achieve.

#2 Design the required capabilities

The second step translates the outcome into business capabilities and jobs to be done. This describes the required outcomes independently of existing organizational hierarchies or technology systems, making the design durable across structural change.

#3 Build the AI-first process

The third step creates an AI activity inventory: a sequenced set of activities that defines, for each step, whether it is handled by AI or by a human. An explicit human-in-the-loop framework governs this distinction – AI is deployed where automation is appropriate; human judgment is preserved where it is critical.

#4 Verify the process logic

The fourth step turns the redesign into a process blueprint designed for implementation. A proof-of-concept (PoC) framework supports stepwise validation before wider rollout.

#5 Integrate change into execution

The final step prepares the redesigned process for scaling across process clusters, supported by governance logic and change integration.

From blueprint to organizational capability

The deliverables from this multi-month project went beyond a single process redesign. The five-step AI-first redesign framework gave the client a reusable logic for future redesign challenges, anchored in a clear outcome statement and success metrics framework that defined what the redesigned process needed to achieve. The AI activity inventory and human-in-the-loop principles translated this logic into a structured template that can be applied across the organization's process landscape.

The impact reflected this shift from one process blueprint to a scalable framework. The client gained full end-to-end transparency on redesigned processes, including responsibilities and dependencies. Automatable steps were clearly distinguished from human-critical ones, while redundant coordination loops were identified so that they could be systematically removed. This created the structural conditions for significant cycle-time reductions.

For automotive organizations facing similar pressures, this engagement offers a replicable AI-first redesign framework – one that includes the five-step approach but goes beyond it by positioning AI not as a tool applied to existing process logic, but as a design principle from which process logic is built.

Results:

Sign up now to receive regular insights on digital transformation and value creation via e-mail.