Abstract
Artificial intelligence (AI) is rapidly transforming the mortgage industry, offering opportunities for enhanced efficiency, improved customer experience and new services. This paper explores how mortgage companies in Europe are leveraging AI across their value chains, provides real-world use cases, and discusses the inherent risks that must be managed to ensure responsible and effective AI adoption.
Introduction: AI applications and potential risks in Europe’s mortgage value chain
The European mortgage industry is traditionally characterised by complex, paper-heavy processes; stringent regulatory requirements; and high operational costs. In recent years, AI technologies—including machine learning (ML), natural language processing (NLP), and robotic process automation (RPA)—have been increasingly adopted by mortgage lenders to streamline operations, enhance risk assessment, and personalise customer engagement. This paper examines key AI applications along the mortgage value chain and addresses associated risks.
AI use cases across the mortgage value chain
As in any industry, AI technologies can potentially have an impact on many areas in the value chain. Alignment between a company’s strategy and its AI use-case road map is crucial to ensure that technological investments directly support the organisation’s long-term goals and competitive positioning. When AI initiatives are closely tied to strategic priorities, such as improving customer experience, increasing operational efficiency, or expanding into new markets or services, they are more likely to deliver measurable value and drive meaningful business outcomes.
Following are some interesting potential use cases for mortgage providers.
Lead generation and customer acquisition
AI, in combination with enrichment by external data, can be used to identify potential borrowers and personalise marketing campaigns in the commercial domain. AI algorithms can segment customers based on behaviour and preferences, enabling targeted outreach and higher conversion rates. For instance, an ML algorithm can be trained to predict clients that are likely to move in the near term, providing interesting leads for mortgage providers and an opportunity to increase market share.
Loan origination and underwriting
In the process of loan origination and underwriting, AI can play a crucial role in different ways. First, AI can be used to assess the completeness and quality of documents that are supplied by a client. Based on this analysis, requests for additional information can be automated.
In addition, AI and ML can be used to assess credit risks by analysing vast amounts of internal and external data. Even when using a rule-based system for underwriting, having a sophisticated AI model as a “challenger model” may help in understanding underwriting risks on a more granular basis. A project done by Milliman in this area, where we developed a challenger model using an ML ensemble method, showed significant potential to reduce default risks.
Property assessment
Computer vision and ML models in combination with external data (e.g., government, property images, comparable sales, market trends) can be used to generate automated property valuations, making the appraisal process faster and more objective. Milliman has built a computer vision model for a Dutch insurer that automatically assesses what type of roof a property has, information that is now used in both their pricing and underwriting models.
Loan processing and servicing
RPA and NLP (by generative AI) can automate routine tasks such as payment processing, customer inquiries, compliance monitoring and message classification. AI-powered chatbots provide 24/7 support for borrowers, while call logging and summarisation using LLMs reduce call centre costs and improve customer satisfaction.
Risk management and fraud detection
AI algorithms are also very useful in detecting anomalies and patterns indicative of fraud, such as document forgery or identity theft. The introduction of (multimodal) generative AI makes generation of fake documents simple, and detecting this type of fraud is an especially daunting task. Research on fake car damages by Milliman has shown that relying on available detection tools alone is not enough, but that a multiple layered line of defence is required.
Prepayment modelling for portfolio projections and valuations
Client behaviour is an important factor for determining future mortgage cash flows. Financial institutions apply various approaches to estimate expected prepayments and the use of the portability option. The level of sophistication varies across financial institutions, but mortgage models typically are based on traditional econometric models such as logistic regression models or options-based models. These consider a limited number of explanatory variables.
ML models can be used to model prepayments, considering many more data features, as well as capture more complex patterns. For instance, the impact of the age of the mortgage taker is expected to have a non-linear impact on prepayments. People tend to relocate during certain stages of their life, and may tend to live shorter in certain locations and types of houses. While this information cannot be used in credit assessments, it may be useful in portfolio projections and valuations.
In addition to modelling prepayments directly, ML models can be used to identify the most important explanatory drivers, which can then be incorporated into traditional models to inform decisions regarding mortgage production. However, our experience replacing regression models with ML approaches for predicting default and prepayment highlighted several challenges, particularly regarding overfitting and unintended interactions among input variables. Successfully implementing such models requires a certain level of expertise and careful consideration of data inputs, as managing these complexities can be more demanding than anticipated.
Clustering and proxy functions
Mortgage portfolios can consist of thousands and up to millions of loans. Often, as part of reporting and risk management, financial institutions value these loans in many different scenarios. This could create a large computational burden. This is especially the case when the Monte Carlo-based stochastic models are used, in which case even a single valuation scenario would require many different evaluations. It is common practice to group loans in buckets with similar characteristics.
AI clustering algorithms can be used to streamline the reporting process and identify the most efficient groupings of contracts. Furthermore, AI can be used to compute proxy functions that capture the value of the mortgage portfolio given risk drivers. Instead of performing a full revaluation of the mortgage portfolio for each scenario, the valuations are determined by the AI-based, less computationally intensive proxy function.
New services
Apart from automating tasks and improving efficiency, AI also offers the potential to develop new services that go beyond traditional lending, creating added value for both customers and the business. One emerging use case is the introduction of AI-powered home affordability and financial wellness tools. These services analyse a customer’s full financial profile, including spending habits, income variability and future earning potential to offer personalised mortgage options and financial planning advice. This helps customers understand not just what they qualify for, but what they can sustainably afford over time.
Another innovative service is instant preapproval and digital mortgage advisors. Leveraging AI-driven document analysis and real-time data validation, companies can provide prospective buyers with near-instantaneous preapproval decisions. Virtual AI advisors, available 24/7 via chat or voice, guide users through complex mortgage processes, answer questions and help compare different loan products tailored to buyers’ unique profiles.
Additionally, AI is enabling proactive loan servicing and retention offerings. For example, predictive analytics can identify when a customer might benefit from refinancing or is at risk of default, prompting timely outreach with relevant solutions. Some companies now offer ongoing property value monitoring and home equity insights as a service, automatically notifying customers of changes that could impact their borrowing power or investment strategy.
These new AI-enabled services not only enhance the customer experience by providing greater transparency, convenience and personalization, but also open new revenue streams and strengthen long-term customer relationships for mortgage providers.
3. AI risks for Europe’s mortgage industry and mitigation strategies
While AI presents significant opportunities for efficiency and innovation in the mortgage industry, it also brings important responsibilities. The use of AI introduces risks such as regulatory compliance and data privacy concerns, algorithmic bias, and operational vulnerabilities. Therefore, it is essential for mortgage companies to implement strong risk management practices. This includes regularly auditing AI systems for fairness and accuracy, safeguarding sensitive customer data, and maintaining transparency in decision-making. By proactively addressing these responsibilities, organizations can harness the benefits of AI while ensuring ethical, compliant and trustworthy operations.
Implementing an AI-model validation approach as developed by Milliman is recommended to ensure AI applications are and remain fit for use, especially for high-risk applications (e.g., credit risk assessment).
4. Conclusion: AI will shape the future of the mortgage industry
AI is fundamentally reshaping the mortgage industry by automating labor-intensive processes, improving the accuracy of risk assessments and enabling the development of innovative customer-centric services. From lead generation and property valuation to risk management and the creation of new digital offerings, AI technologies are driving significant gains in operational efficiency and competitive differentiation. However, the rapid adoption of AI also introduces new risks, particularly regarding data privacy, regulatory compliance and algorithmic fairness. To fully realise the transformative potential of AI while maintaining trust and integrity, mortgage providers must embed robust risk management practices, ensure transparency and regularly validate AI systems. As the industry continues to evolve, organizations that strategically align their AI initiatives with business goals and proactively address emerging risks will be best positioned to deliver sustainable value to both their customers and stakeholders.