Data Readiness: Why Only 7% of Swiss Firms Can Scale AI

57% of Swiss companies now use AI in some form, up from 46% a year earlier. Yet only 7% rate their data as fully ready to generate systematic AI-driven value. The bottleneck for the next stage of scaling is no longer tools, budget, or model choice — it is data quality and structure. That is the central finding of 'Unlocking the Potential of AI in Switzerland 2026,' a study by AWS and Strand Partners based on roughly 2,000 respondents. The same pattern holds globally: a study by Cloudera and Harvard Business Review Analytic Services (March 2026) among more than 230 executives comes to the exact same figure — worldwide, only 7% of enterprises say their data is completely ready for AI, with a further 27% rating it as not very or not at all ready.

Why the Bottleneck Has Shifted

Two years ago, most conversations centred on budget, tool selection, and which model to choose. Those questions are largely settled now. What remains is less comfortable: 55% of companies describe their data as only 'partially ready,' and a further 22% as 'largely ready' — leaving a small group with a data foundation solid enough to carry AI value creation at scale. This aligns with the Swiss AI Observatory 2026 from Colombus Consulting: 82% of surveyed companies rate their data and AI maturity as low to medium, and the share rating their own data quality as good or excellent has dropped from 61% to 51% — not because the data got worse, but because the bar each new AI project sets keeps rising.

The Computerworld Swiss IT Study 2026 puts concrete numbers on this: 32.1% of companies cite data quality as the biggest barrier to AI use cases — more than double the share citing budget or management support. The adesso GenAI Impact Report 2026 deepens the picture: the largest hurdles are poor data quality (32%) and missing processes and governance (32%) — both ahead of lack of staff qualifications (30%) and security concerns (29%). Strikingly, none of these barriers are technological in nature. Hallucinations, often perceived as the quintessential AI problem, rank only ninth at 18%. The bottleneck is organisational, not technical.

Three Levels of AI Maturity Among Swiss Companies

• Basic level (60%): chatbots and off-the-shelf tools without deeper integration into processes or systems.
• Intermediate level (22%): AI is integrated into existing workflows and data sources but does not yet operate independently.
• Advanced level (18%, down from 20% a year earlier): agent-based approaches and combinations of multiple models — a share that has slightly declined even as expectations for this level have risen.

Agentic AI Without a Foundation: The Risk for Early Movers

8% of Swiss companies already run agentic workflows — a share above the global average. That sounds like an edge, but it cuts both ways: an agent operating on incomplete or inconsistent data amplifies errors rather than avoiding them. Why existing governance frameworks structurally fail in precisely this scenario is something we unpacked in our analysis of the AI Agent Governance Gap.

Why Companies Keep Investing Anyway

Low data maturity is not holding Swiss companies back from investing. Nearly 30% expect a strong increase in AI spending — twice the global average. 65% assign AI high strategic priority, and 70% consider it central to their company strategy. This combination of high ambition and low data readiness is not a contradiction; it is a pattern that shows up across markets where AI adoption outruns the data foundation meant to support it.

Data Structure Before Tool Choice: The Real First Step

On the Swiss AI Podcast, Chris Jon Graf puts it plainly: the success of AI projects is not decided by which tool you choose, but by data structure and process readiness. Reverse that order and you are buying disappointment, regardless of how capable the chosen model is.

The question is never which model we use first. The question is whether our data and processes are even capable of carrying what we expect from that model.

This gap between pilot and production shows up across industries: companies score early wins in pilot projects but stall at the transition to production because the data foundation cannot carry the jump. Recognising that pattern early is what separates companies that scale AI from those that keep restarting pilots — something we examined in depth in our guide on escaping the pilot trap.

What This Means for Swiss Decision-Makers

The evidence is clear: AI adoption in Switzerland is growing faster than the data foundation meant to support it. For decision-makers, that does not mean slowing investment plans — it means clarifying the sequence before the next spending round begins. Knowing your real data maturity also tells you which of the three levels is the next achievable step. This aligns with our analysis of the Swiss SME deployment race: it is not model benchmarking that determines success, but the ability to integrate AI into real processes. That is precisely where we start with our clients: not with model selection, but with the question of whether the data can carry what the business expects from AI.