Real results from real companies

The potential analysis revealed that the biggest time losses are not caused by individual "big problems" but by many small friction losses that accumulate throughout the day. Product knowledge was spread across documents, lists, and experiential knowledge, meaning relevant information often had to be searched for and compiled first. Additionally, inquiries were frequently incomplete, leading to follow-up queries and multiple coordination rounds before a solution could even be developed. In warranty and claims cases, recurring check steps also emerged that were predominantly run manually and documented differently depending on the processor.

From these findings, a prioritized roadmap was derived that addresses the biggest time drains first. This includes an internal "AI product advisor" that guides employees step by step to the right article number and optionally verifies the result through a brief human approval, as well as an intelligent search that makes relevant passages from manuals and lists immediately accessible. Additionally, building blocks were defined to relieve communication – such as automated pre-checks for complaints (completeness, plausibility, next step), email pre-sorting with routing to the right department, and support for requesting missing information through clear, automatically prepared follow-up queries.

The potential analysis revealed that major effort in office work arises especially where information is not cleanly available in one place. Employees had to consolidate data from different sources, reconcile, and transfer to ERP – a process that not only costs time but is also prone to typos or misunderstandings. Particularly challenging were differently structured documents and missing mandatory fields that triggered additional queries. At the same time, it became clear: for automation to work, transparency is needed – the ability to trace what was automatically recognized and where human review remains sensible.

Based on these findings, a roadmap was defined that leads step by step from manual entry to semi-automated and later largely automated workflows. A central building block is a standardized handover to ERP so that automatically captured data can be reliably processed. Building on this, AI assistants were planned to structurally extract content from email text as well as make information from attachments – including scans via OCR where needed – usable. The whole approach is complemented by quality and monitoring functions that make visible where automation already runs stably, where uncertainties occur, and where adjustments should be made.

The potential analysis showed that the biggest friction losses resulted not from missing technology but from inconsistent usage and missing standards. Information was distributed, responsibilities were not always clear, and this led to search times, duplicate filing, and unnecessary queries. Email inboxes were partly used as filing systems, making retrieval and traceability difficult. Additionally, it became visible that knowledge about norms and standards existed but was not quickly enough accessible in daily work – meaning employees had to research again and again and decisions took longer.

Instead of a large IT project, a pragmatic implementation path was developed that can deliver results quickly. At the center were clear rules and a clean organizational system: an unambiguous inbox structure, fixed filing and naming rules, clear responsibilities, and a short rollout with test phase, onboarding, and subsequent refinement. As an optional next step, an AI assistance system was outlined that can provide quick orientation on norms and quality questions – understandable, traceable, and designed to actually be used in daily work.