AI in the Enterprise: Why Knowledge Management Decides Between Success and Failure
Christian Kirsch in the AFS magazine: in mid-market companies, AI rarely fails because of the technology – it fails on data quality, permissions, context and governance. A guest article with a pragmatic starting point.
Our managing director Christian Kirsch has published a guest article in the AFS Academy magazine – on a question that comes up in almost every mid-market AI project we run: why does AI so often fail when the technology has long been available?
His thesis is uncomfortable but honest: AI rarely fails in a company because the model is too dumb. Most of the time it fails because the company itself isn’t ready yet. The real foundation isn’t the model – it’s the knowledge management behind it: structured data, clear access rights, context and governance.
Four reasons AI fails in mid-market companies
In the article, Christian spells out the typical pitfalls:
- Poor data quality – AI needs findable, up-to-date and well-maintained information.
- Unclear permissions – who is allowed to see what? Without an answer, every AI becomes a risk.
- Missing context – data without context delivers arbitrary, unreliable answers.
- Lack of governance – without ground rules, adoption stays a matter of chance rather than a system.
Start pragmatically instead of waiting for the grand plan
Rather than starting with hype use cases, Christian argues for a sober entry point: begin with a concrete problem, build on the existing work environment – for many companies that’s Microsoft 365 – and plug AI in where it noticeably takes work off people’s plates. Because a system only gets used when it genuinely makes everyday work easier.
This very stance – digitalisation as a craft, not a slide topic – runs through our AI consulting.