From Data to Decisions: AI along the Mortgage Value Chain
A Personal Review of the Hypoforum 2025
On November 5, I had the opportunity to give a presentation together with Thomas Zerndt, CEO of the Business Engineering Institute St. Gallen, at the 17th Hypoforum in Zurich. Our topic was the targeted use of Artificial Intelligence along the mortgage value chain. The title of our presentation, “From Data to Decisions”, captures the core question: How can AI not only automate processes but also create real added value in customer interaction and in the quality of decisions?
Four stages, many possibilities
The lifecycle of a mortgage can, in simplified form, be summarized in four steps, from the initial property search to the final credit decision. Along this chain, each stage offers concrete and in some cases already tested application areas for the use of AI.
1. Property search
For many, property searches begin with standardized filter functions on familiar platforms, but AI opens new paths. Even today, I can simply state as a user: “I am looking for a quiet, bright apartment with a balcony within biking distance of Zurich HB” and an AI system will understand my needs and deliver suitable suggestions. This is exactly what large language models such as ChatGPT enable.
A scenario in which buyers and sellers are connected directly through a dialog-based system that is personalized, efficient, and intuitive (see Figure 1, illustration of a match between buyer and seller on ChatGPT) could be realized in the future. In addition, such systems could one day suggest potential properties that are not yet on the market but are likely to be sold soon, based on behavioral data, historical transactions, and market developments.
Of course, this raises questions about data protection, traceability, and the acceptance of such predictions. It remains open whether and when such solutions will prevail, but the technological direction is clear and compelling.
2. Mortgage Application
The mortgage application process is often accompanied by a flood of documents, which is a hurdle for both customers and banks. This is exactly where AI-based agent systems come in: they automatically read submitted documents, check their completeness, and extract relevant data in structured form.
An example of a modular agent system (see Figure 2, simplified representation of the automated AWS mortgage process), which works according to the principle of Amazon Web Services (see link aws), shows how the application process can be automated. A central supervisor agent controls the overall process and coordinates specialized sub-agents for document recognition, validation, compliance checks, and data transfer. The result is shorter processing times, fewer manual errors, and significantly more transparency.
3. Property valuation
Property valuation was long a static process that relied heavily on hedonic models. With AI, especially neural networks, satellite data, and reinforcement learning, it is now possible to build models that are more dynamic and context sensitive.
Particularly exciting is the use of generative AI for scenario analyses: What happens to a property’s value after renovation? Or if the neighborhood is upgraded due to new infrastructure or socioeconomic changes? Such simulations can be realized with so-called digital twins, meaning digital copies of real properties on which various construction or market scenarios can be tested.
For example, based on historical market reactions, one can simulate which renovation variant promises the greatest increase in value (see Figure 3, simulations of different scenarios with a digital twin). Future collateralization could also be designed more dynamically, for instance when AI-based market development simulations anticipate positive value trends (see Figure 4, digital twin and increased collateralization based on it). The potential is large, while at the same time the use of data-driven diligence is necessary to avoid over-optimizations or model errors.
4. Credit Decision
The final credit decision is increasingly being shifted toward real time through AI. Applications can be evaluated in a risk-adjusted way based on current market data, customer behavior, and ESG criteria, with immediate influence on pricing and conditions.
The central challenge is that these decisions must be not only efficient but also transparent. Explainable AI (XAI) therefore becomes a key concept. Neither customers nor advisors, much less regulators, accept decisions whose origins cannot be traced and verified.
Especially in sensitive areas such as mortgages, clear governance, monitoring, and a deep understanding of model functionality are essential. The principle of the human in the loop is crucial here: human experts remain actively involved in critical decision processes to check the plausibility of results and intervene if necessary. This is the only way to maintain trust and use the advantages of AI responsibly and in the interest of the customer.
AI is more than technology
The most important insight from this day is that AI is no longer discussed as a future topic but as a fixed part of everyday banking. Many participants reported ongoing pilot projects and initial productive implementations. At the same time, it became clear that without ethical guidelines, clear governance structures, and clean data quality, AI cannot unfold its potential.
Conclusion
Artificial Intelligence offers enormous transformation potential along the entire mortgage value chain. Whether in property search, application processing, property valuation, or credit decisions, AI can not only accelerate and streamline processes but also improve them qualitatively.
What matters is a responsible approach to these technologies. Explainable models, robust data quality, and clear regulatory frameworks form the foundation. In addition, the principle of the human in the loop plays a central role, since in critical phases, human judgment remains indispensable.
Those who manage to combine technological innovation with trust, transparency, and customer value can help shape the mortgage world of the future.
I look forward to continuing the discussion and to developing concrete solutions together with partners.
Sources
Figure 1 Own illustration of a match between buyer and seller on ChatGPT
Figure 2 Own simplified representation of the automated AWS mortgage process
Figure 3 Simulations of different scenarios with digital twin, own illustration
Figure 4 Digital Twin and increase of collateralization based on it, own illustration
With an academic background in mathematics (MSc, University of Zurich) and professional experience as an IT consultant in Zurich, he combines analytical strength with practical implementation. In various projects at major Swiss banks, he was responsible for the development and implementation of solutions in the areas of data migration, reconciliation, and automation. He used Python and SQL to optimize complex processes and make data usable as a measurable value for the business.
Through a combination of research, teaching, and consulting, he is working to design scalable, data-driven solutions for the banking of the future.