Developing AI innovations in a corporate environment in a pragmatic, user-friendly, and scalable way is a challenge. In this interview, Frank Mielke, Development Lead at Bosch Digital, explains how his team implements new solutions at high speed, why proximity to users is a strategic advantage, and how the transition from GenAI to Agentic AI can be conceptually anchored from the outset.
Mr. Mielke, how does a high-speed team strike a balance between "deliver quickly now" and "build for global sustainability"?
Even the most agile innovation or AI team within a corporation always has to contend with complex processes. It is important to be aware of this duality. Rollout usually takes the most time. For example, we rolled out the internal AI assistant “AskBosch” across the group for 336,000 employees with IT access. That simply takes time, especially considering the many different business units. This makes speed in development all the more important. We work in very fast cycles here. Our approach is to build on proven best practices, for example when it comes to user interfaces. We look to market leaders to deliver solutions that work quickly. We pick out what works and improve it in areas that are crucial for users.
Where do you start when identifying these decisive levers from a user perspective?
The question is: Where can we add the most value? We have the opportunity to develop in a user-centric way with as little overhead as possible. Good requirements and feedback management is key to prioritizing and implementing the right topics at the necessary speed. And: measurability. Currently, this is often only done on the basis of anecdotes, but it needs to be addressed structurally. User feedback is also crucial here. It can be used to track which functions are being used, for example.
You mentioned earlier that AskBosch is already available to over 300,000 employees. How can you ensure that AI is used responsibly?
Education is the gateway. An example: Before using our internal assistant "AskBosch" for the first time, users are presented with the guidelines and must take a short quiz to prove that they have read and understood them. Once they have gotten started, it is important to guide users and incorporate moderation and transparency into the design of the user experience. For example, if employees leave the secure sandbox during grounding and switch to the official Google search, they receive a corresponding message informing them that sensitive data must not be entered at this point.
The full interview with Frank Mielke can be found in our whitepaper "Road to Agentic AI. The fascination of automation."