Unlocking the future: Artificial intelligence–powered horizon scanning for financial stability
Abstract
Financial supervisory and regulatory organisations increasingly operate under conditions characterised by heightened uncertainty, a higher incidence of large systemic disruptions, and rapid socioeconomic transformation. In this context, horizon scanning has emerged as an institutional strategic intelligence practice for systematically detecting early signals as indicators of potential change. Despite its growing adoption, horizon scanning remains largely descriptive, episodic and weakly operationalised, limiting its standardisation and effectiveness for continuous monitoring and anticipatory decision-making in high-stakes environments such as finance. In this paper, we advance horizon scanning from a predominantly conceptual and qualitative episodic practice, towards a resilient, computationally grounded, and institutionally governable capability designed with, but not limited to, a suitability for financial supervisory and regulatory use in mind. We inform a design and engineering blueprint for a supervisory and regulatory grade horizon scanning computational system based on modular decomposition of cross-domain drivers of change and their recombination into structured, semantically plausible yet novelty enabling future scenarios. The proposed framework leverages agentic, generative artificial intelligence (AI) to support continuous horizon mapping, exploratory scenario generation, and policy-oriented evaluation, while embedding human-in-the-loop oversight, auditability, and ethical governance mechanisms. The paper introduces agentic generation and evaluation of modular integrated scenarios as a concrete instantiation of this design logic and discusses its functional value for central banks and financial supervisory and regulatory organisations. The paper concludes that modular, agentic horizon scanning has the potential to enhance institutional anticipatory capacity and support resilient policy design under uncertainty, while expert human-in-the-loop provides a robust governance foundation for the responsible use of AI in strategic foresight. This paper is also included in The Business & Management Collection, which can be accessed at https://hstalks .com/business/.
The full article is available to subscribers to the journal.
Author's Biography
Lucia Gomez is a professor of applied data science and finance at Bern University of Applied Sciences, Switzerland and senior researcher and lecturer at the University of Geneva, Switzerland. With a background in psychology and a PhD in computational neuroscience — and, therefore, a major interest in human-in-the-loop technology — she specialises in the development of interdisciplinary artificial intelligence (AI) solutions with a dedicated focus on generative AI. The Institute of Applied Data Science and Finance of the Bern University of Applied Sciences is a leading Swiss research institute for data-driven, finance-based and strategic insights, analysis and value creation. The Geneva School of Economics and Management at the University of Geneva is a leading Swiss Business School, with a mission to educate responsible leaders for a diverse and changing society and serves as a leading research centre focusing on analytics/digital, governance, statistics and information systems.
Branka Hadji Misheva is a professor of applied data science and finance at Bern University of Applied Sciences, Switzerland, where she leads the Applied AI Research and Solutions group. She has led multiple research projects on AI applications in finance, with a particular emphasis on credit risk modelling and explainability in machine learning. Her research centres on explainable AI, network-based models and FinTech risk management, with a strong focus on developing transparent and trustworthy AI systems for high-stakes financial decision making.
Thomas Maillart is a senior lecturer at the Geneva School of Economics and Management, Switzerland, specialising in complex techno-social systems and the dynamics of collective intelligence. His work involves building quantitative models to understand human interactions with systems such as the internet through innovations. Holding a PhD from Eidgenössische Technische Hochschule Zurich, Maillart has extensively researched the application of collective intelligence in education and innovation.
