A representative survey of 603 German companies marks a threshold: for the first time, a majority use AI. At the same time, most users say they barely tap its potential, and only a few see themselves as intensive users.
The cost structure is revealing. Infrastructure, data preparation and integration with existing systems are named as a major cost far more often than licences, which only a fifth of users see that way. This confirms what we see in ERP projects: contradictory master data and parallel spreadsheets prevent reliable results, whichever model is used.
It is also striking where AI works today. Customer contact and communication dominate, while it is far less common in production and accounting. For mid-sized companies, it therefore pays to budget for data quality from the start and to choose the first use case deliberately in a core process such as procurement or service.