Insights from the Roundtable

25 Feb 2026

The Swiss AI Summit Community (SWAIS Community) kicked off the year by tackling the question: “How to create Real Business Value from AI”? The roundtable discussion featured Jan Mikolon, CTO of QuantumBasel, and Arlind Spahija, Head AI of Migros Bank.

The roundtable brought together board members, executives, and senior practitioners to discuss why artificial intelligence initiatives so often fail to scale and what leaders can do to turn experimentation into sustained business value. While enthusiasm for generative and agentic AI is high, participants agreed that governance, data readiness, and human factors remain decisive constraints.

Why 82% AI Pilots Fail to Scale

Participants repeatedly referenced the widely observed phenomenon that the majority of proofs of concept never reach production. The reasons are rarely technical alone.

What are the reasons for the high failure rate?

- unrealistic expectations about what AI can deliver,

- insufficient attention to data quality, and the

- tendency to treat AI initiatives as isolated experiments rather than as part of a broader transformation agenda.

A recurring theme was that AI is frequently layered onto inefficient or poorly designed processes. In such cases, AI merely accelerates existing inefficiencies instead of creating value. Without clear ownership and accountability across end-to-end processes, even technically successful pilots struggle to survive organizational reality.

Measuring Business Value Beyond Experimentation

A central discussion point concerned the difficulty organizations face in measuring AI value. Many initiatives begin without a shared understanding of the intended business outcome. The roundtable converged on the view that value must be assessed through a combination of usecase impact, adoption rate, scalability, and total cost of ownership.

Time savings alone were seen as an incomplete metric. Participants emphasised timetomarket, quality improvements, and the ability for employees to focus on highervalue work as more meaningful indicators. Comparing AIsupported performance with humanonly baselines was highlighted as a practical way to demonstrate incremental value.

Data as the Hidden Bottleneck

While AI is often discussed in terms of models and tools, the discussion made clear that data engineering and data quality dominate realworld effort. In several industries, the majority of project time is spent preparing, cleaning, and integrating data before AI can be applied at all. Poor data quality not only delays projects but also undermines trust in outcomes.

This reality differentiates AI initiatives from many traditional IT projects.

Leaders must recognize that data readiness is a strategic investment rather than a technical afterthought.

Human and Organizational Factors

Beyond technology and data, human factors were identified as equally critical. Fear of job displacement, lack of AI literacy, and misaligned incentives can all hinder adoption.

Participants stressed the importance of involving end users early, framing AI as a productivity multiplier rather than a costcutting tool, and investing in education at board, executive, and employee levels.

A culture that allows learning from failure, rather than penalising it, was seen as essential. Without psychological safety and clear communication, even welldesigned solutions face resistance.

From Everyday AI to Customized AI

The discussion distinguished between three broad categories of AI adoption. First, everyday AI tools that improve individual productivity through standard, offtheshelf solutions. Second, domainspecific AI tailored to particular functions such as legal, research, or industrial applications. Third, fully customised AI built on proprietary data and integrated into core processes.

Participants agreed that while everyday AI delivers quick wins, sustainable competitive advantage increasingly depends on customised AI aligned with strategic priorities and governed at the highest levels.

Implications for Boards and Executives

The roundtable concluded that AI is no longer an IT topic but a leadership responsibility. Boards and executive teams must be sufficiently literate to challenge assumptions, prioritise the right use cases, and oversee risks and investments. Moving from pilots to production requires clarity of purpose, disciplined governance, and sustained commitment across the organisation.

AI, when treated as an organisational transformation rather than a standalone technology project, can deliver measurable business value and strategic resilience.

Key takeaways

• AI must be tied to business outcomes. Successful initiatives start with clear business problems and measurable impact rather than technology experimentation.

• Most AI projects fail to reach production. Up to 85% of AI initiatives stall at the pilot stage, often due to poor alignment with business needs, unrealistic scope, or weak execution.

• Focus on practical use cases that deliver measurable value. Real-world implementations—such as internal assistants, email automation, and knowledge retrieval systems—can significantly improve efficiency when integrated into daily workflows.

• Strategic alignment between business and technology is critical. AI initiatives succeed when leadership, business teams, and technical teams collaborate closely from the beginning.

• Learn from proven implementations. Case studies from organizations like Migros Bank highlight the importance of clear priorities, iterative deployment, and transparent evaluation of results.

• Not every problem requires AI. Validating feasibility and impact early helps organizations identify where AI truly creates value—and where it does not.

• Interactive validation accelerates adoption. Testing real use cases against organizational challenges helps leaders move from ideas to actionable next steps.