German companies buying AI are increasingly buying an implementation problem with a software contract attached.
That is not a criticism of the technology. It is what happens when a tool moves from individual experimentation into an industrial or enterprise workflow.
Digital maturity determines how much value a buyer can absorb
KfW's Mittelstand research repeatedly finds that AI adoption is associated with existing digital capability, innovation activity and know-how. Companies with stronger foundations are better placed to integrate new systems.
A vendor can supply a capable model, but it cannot instantly repair fragmented data, unclear process ownership or an ageing software estate.
Procurement therefore needs technical questions
Who owns the data? Where does inference occur? Which systems need to be connected? What happens when the model is wrong? How are permissions managed? Can the customer change provider without rebuilding the workflow?
Those questions belong in procurement because they shape long-term cost and risk.
The Mittelstand raises the bar for practical ROI
Large multinationals can fund broad experimentation. A mid-sized manufacturer usually needs a narrower business case. The value proposition has to connect to throughput, quality, labour, maintenance, sales or another measurable operating outcome.
This can favour specialist vendors over broad platforms when the specialist understands the process deeply. It can favour large platforms when integration and security matter more than marginal model performance.
Buyers should compare implementation evidence
Case studies are most useful when they identify the baseline, deployment scope, cost, measurable result and limits. German AI procurement is becoming less about whether a product contains AI and more about whether the organisation can use it reliably.