Between the loud AI promise and quiet execution lies a course correction that pragmatic voices like Niklas Steenfatt have long called for: away from hype, toward measurable impact. For SMEs this is the decisive lesson, because nowhere is the gap between excitement and results as expensive as here.
What is the pragmatic AI course correction?
The core is simple: do not introduce AI as an end in itself, but as a tool for a clearly named bottleneck. That sounds unspectacular, yet it contradicts the common reflex to start as many initiatives as possible at once. A widely cited 2025 MIT study shows why that reflex is costly: around 95 percent of enterprise AI pilots deliver no measurable effect on the bottom line. The pragmatic approach flips the logic and starts with the problem, not the tool.
What can SMEs learn from it?
Three things. First, patience with the J-curve. New technologies cost time and money before the return kicks in, so expectations must be realistic. Second, buying often beats building. Specialized, purchased solutions reach productive use in about two out of three cases, internally built general-purpose tools only a third as often. Third, measure instead of opine. Without a metric, every AI initiative is a matter of taste.
What does pragmatic AI look like day to day?
Pragmatic AI is boring and effective at once. It sits on a recurring process, has an owner, a metric and a stop criterion. The human keeps authority over decisions, the AI handles the legwork. This sober approach turns a trend topic into a contribution to the bottom line. Whoever works this way does not need the hype, because the results speak for themselves.
Further reading
These articles help with the next decision.
- Why AI Understanding Often Fails in SMEs
- The Overlooked AI Revolution in Germany
- AI, ML and Deep Learning: Basics from Practice
What does a pragmatic AI course correction mean?
Introducing AI not as an end in itself but as a tool for a clearly named bottleneck. The approach starts with the problem and the metric, not the tool.
Why do many AI initiatives fail?
Because too many projects start at once without prioritization. Around 95 percent of pilots therefore deliver no measurable effect on the bottom line.
Buy or build?
For most SMEs buying is the safer path. Specialized solutions reach productive use about twice as often as internally built general-purpose tools.
What does pragmatic AI look like day to day?
A recurring process with an owner, a metric and a stop criterion. The human decides, the AI handles the routine.