The implementation of Chatbots and Robo-advice in the mortgage market

Abstract

This Master’s thesis at the Delft University of Technology is executed in collaboration with Fintech company Yellowtail.

The assignment was to design a new mortgage application process. The current consumer journey proved to be outdated and a shocking mismatch with the new generation of firsttime buyers. Especially the high costs and the obligation to consult an adviser made the process almost a burden to buy a house. Before any
designs were made, the context is explored. The chosen approach consisted of a
quantitative and qualitative study. The former revealed the lack of financial knowledge, especially of debts. It also showed the insecurity of consumers and a wrong perception of what the advisers do. Many first-time buyers often consult
family and friends but these groups did not seem to be the most reliable regarding their knowledge.
In the qualitative study, 5 participants were interviewed who just bought their first house, two of them through an execution only channel. Based on these interviews, the current mortgage application journey is made. Based on the
transcripts, 8 codes were identified and placed onto this journey map. The codes were used as needs of the consumer. It exposed which emotions had an effect in which stage of the process. From the qualitative study, two persona’s were created. Together they determine the target group of this project. The design brief in combination with a new journey map showed what had to be designed in the second part of the report.

Firstly, a model is made transforming user input into a mortgage advice recommending the best mortgage amount, fixed interest rate and amortization type. Different variables which have an effect on the three advice points are analysed.
The model consisted of multiple calculations and boundary conditions. It is an invisible model for the consumer, only the output is shown to them. Secondly, the front end of the system is created in the interaction design chapter. Several ideas were created and tested with consumers in a small user test. Creating overview proved to be more difficult than expected. Also many consumers still had the need of an adviser. Although this might change within a few years
with the rapidly evolving artificial intelligence technology, the concept in this report starts with a small role for the adviser. This person will check the application and guide the consumer to the notary.

Lastly, a chatbot conversation is written. It is presented as an addition to the interaction design. Literature explained several steps to create a convenient and trustworthy chatbot. A small user study is conducted to determine the best
tone of voice. The three elements: the model, the interaction design and the chatbot together form the final design. For technical reasons and limited time they are presented independently.

The reports ends with a view on the future and the role of the advisers.