AI rarely fails on the technology. It usually fails on the foundation.

Gartner estimates (June 2025) that over 40 percent of agentic AI projects will be scrapped by the end of 2027. The cause is rarely the tool. It's the business the tool lands in.

Why AI projects fail in practice

An SME buys an AI tool — and three months later nobody uses it. Not because the model is too weak. But because the process it should support lives only in people's heads; because the data it would need is scattered and outdated; because the core system won't give up its data; and because nobody clarified what may go in at all. Four pillars decide — not the model.

Processes and organisation: why only documented work can be delegated

«Ms Müller handles that — she knows how it works.» That very sentence is the most common AI blocker. Only documented processes can be delegated — to people and to AI alike. Five lines often suffice: trigger, steps, who decides, where it snags.

Data hygiene: why AI is only as good as the data it reaches

AI is only as good as the data it can reach. Scattered, outdated repositories are the quietest blocker — no tool fixes them. Even a simple map of your key data holdings shows where tidying up pays more than buying.

Infrastructure: when an old system is ready for AI

The decisive question about your software isn't how old it is — it's whether it gives its data back. Export or interface available: connectable. Neither: an island — and every connection gets expensive.

Compliance: what to settle before the first rollout

«Are we even allowed to do this?» The honest answer starts with three counter-questions:

• What data flows in?
• Is there a data processing agreement with the provider?
• Who in your company answers access requests?

Open points are no showstopper, they're the list for your specialist.

Where you start

With an honest assessment: strongest pillar, biggest gap, first step. That's all it takes to carry decisions — honesty beats completeness. We work it out with you in the AI audit. Or do it yourself: “The Groundwork”, module 2 of our workplace AI training, takes you exactly there in four short working sessions.

Tools age. Your foundation doesn't.

We guide Swiss SMEs along this path — human, competent, holistic.