Developing an AI Decision Engine for Bias-Free Lending
We developed an AI model to add objective decision-making to the mortgage loan approval process, improving fairness and transparency.
About the client
Wirtschaftsagentur Wien is dedicated to strengthening and promoting Vienna as a business location on behalf of the City of Vienna. They support ventures that advance sustainability, innovation, quality of life and diversity with funding and advisory.
With its Innovation and Employment funding, waff supports small and medium-sized enterprises in Vienna in the implementation of innovation projects.
To make mortgage lending fairer and more transparent, we teamed up with a leading bank to build an explainable AI layer on top of 50 000+ historic data points from loan applications. The result: faster approvals, measurable bias reduction and regulator-ready transparency.
At a Glance
Objective
Enable fair, transparent, data-driven mortgage approvals by reducing bias and increasing trust through explainable AI.
Solution
Trained an explainable model with 50k+ data points from past applications, integrated it into the bank’s workflow and visualized in a dashboard that shows the weight of every factor in real time.
Result
We were able to reduce time-to-yes for mortgage loans significantly. Also, the quality of decision making improved due to the availability of additional information and learnings from the past.
Our project focuses on enhancing the mortgage loan approval process at banks by integrating an AI-based objective decision-making layer. We began by collecting and structuring over 50,000 datapoints relevant to mortgage applications. This comprehensive dataset underwent extensive data exploration and analysis to uncover patterns and insights critical to the lending decision process.
Using these insights, we developed an AI model designed to augment current decision-making frameworks with precision and fairness. To make the AI's decisions transparent and explainable, we created a detailed dashboard that visualizes the decision-making process. This dashboard highlights the weights assigned to each parameter, ensuring that users can see how various factors influence the AI's recommendations.
Furthermore, to maximize the utility of our data, we incorporated a Large Language Model (LLM). This easier to extract meaningful insights and make informed decisions. The integration of LLM ensures that the wealth of data can be accessed and utilized efficiently, facilitating a deeper understanding and better-informed decision-making process.
Overall, our project not only streamlines and enhances the mortgage approval process but also ensures transparency, fairness, and ease of use, paving the way for more reliable and equitable lending practices.
Summary
About the project
Our project enhances mortgage loan approvals by integrating an AI-driven decision-making layer. We analyzed 50,000+ data points to develop a fair, precise model with a transparent dashboard that explains the AI's decisions. A Large Language Model (LLM) was also added to extract insights efficiently, making the process more transparent, fair, and user-friendly.
Services provided
Visual Design
Definition of Target Architectures
Design of IT Architectures
Front-End and Back-End Development
IT Security Consultation
Cloud or On-Premise Hosting
Operate and Maintain Platforms
Support to Scale Systems for Future Growth
Like how we work?
You’re just one email away from creating a partnership you won’t regret. Send a few lines to hello@viable.at and let us know what you’re up to and where we can help.