From Pilots to Production: Scaling AI in Swiss Finance

Artificial intelligence (AI) is rapidly reshaping our lives. Yet, despite growing investment and enthusiasm, most AI initiatives in the Swiss financial services industry remain stuck in the proof-of-concept (PoC) phase. A new study by Swiss FinTech Innovations (SFTI) and the Institute for Finance & Law at the Eastern Switzerland University of Applied Sciences (OST), supported by ELCA Advisory, investigates why – and how institutions can close this “AI deployment gap.” 

The study, AI from PoC to Production – Bridging the Deployment Gap in Swiss Financial Services, is based on a survey of SFTI members and in-depth expert interviews. It reveals an interesting paradox: while the number of AI pilots has increased sharply, the share of use cases in full production has fallen compared to our  2024 study. In 2025, about 40% of all AI initiatives are still in ideation or feasibility analysis, while only 17% have reached deployment and 11% scaling (see Figure 1). 

Fewer, better-prepared deployments 

This shift is not necessarily negative. Institutions are becoming more selective, focusing on fewer but better-prepared initiatives. The study found that 65% of production cases met or exceeded expectations – a strong indicator that with disciplined governance and robust preparation AI’s potential can be successfully productized. “Quality clearly beats quantity,” as one participant put it. 

The key differentiator between stalled pilots and scaled production lies in what the study calls institutional readiness: the ability to align data quality, governance, and business objectives. Institutions with well-defined business cases, strong data foundations, and compliance-by-design processes move faster and deliver higher impact and business value. 

Data and governance at the core 

Data privacy, security, and quality remain the top barriers to scaling. Fragmented data landscapes, coupled with complex approval processes, often slow progress. Meanwhile, regulation plays a dual role: it adds complexity but also creates trust. FINMA’s Guidance 08/2024 on AI governance has set a clear direction, emphasizing accountability, transparency, and explainability. 

The study identifies four dimensions that determine success:  

• strategic alignment 
• data and MLOps infrastructure 
• governance and compliance 
• adoption and scaling  

Together, they form a holistic framework supported by seven practical action steps – from “production-first design” to “portfolio governance” – as laid out in the study. This structure helps institutions prioritize foundational investments before expanding their AI portfolios. 

From principles to practice 

The recurring theme of discussions with industry experts during the ‘Digital Finance Day 2025’ was, success depends on operationalizing and embedding your AI strategy. Rather than developing isolated pilots, institutions should design AI use cases with clear deployment criteria, measurable KPIs, and early compliance involvement. As one participant noted, “A PoC should be the first step of production, not an experiment on the side.” 

Collaboration as a catalyst 

Finally, the study underscores the importance of collaboration. Through platforms such as SFTI, banks can develop shared standards – for governance templates, success checklists, and maturity assessments – reducing duplication and accelerating progress across the ecosystem. 

The conclusion is clear: AI in Swiss financial services has entered a phase of consolidation. The challenge is no longer figuring out what AI can do, but ensuring AI delivers on the promised benefits: securely, measurably, and at scale. 

Full study: AI from PoC to Production – Bridging the Deployment Gap in Swiss Financial Services, Available at: https://swissfintechinnovations.ch/projects/bridging-the-ai-poc-production-gap-keys-to-deployment-success-in-swiss-financial-industry/