"Collette" -  the Mortgage Adviser of the Future

"Collette" – the Mortgage Adviser of the Future

mortgage market view or MMR was an update to the mortgage rulebook that took effect from April 2014 it meant that most customers now have to receive advice before they can apply for a mortgage this was a major challenge for the industry the IDB I collect was to see how far we can go with artificial intelligence to create a virtual mortgage crisis so the Innovation Program has always been about taking the latest technologies and applying them in new and interesting ways for our clients and artificial intelligence is of particular interest to us and is really the technology that is behind the Robo mortgage advisor the entry point for our prototype is the agreement in principle page we've assumed the customer has already gone through the credit and affordability checks in order to get to this stage this means that we already know some basic information about the customer and the property they're looking to by clicking chat now launches the advice conversation with Colet this takes the form of a web-based text chat Colet asks a series of questions covering topics like fees term length rate type and overpayments each time a customer makes a choice Colet asks them to explain their decision to ensure the customer has fully understood the key concept and the implications of making that choice throughout the advice conversation Colet gathers the customers choices and uses these to provide a tailor-made recommendation Colet will compare the information she's collected to a list of mortgage products and will present a product that fits the customer's needs and circumstances is able to consider the same advice areas that normal human adviser would do it works in exactly the same way as a normal interview of a mortgage advisor in that a fact finders undertaken to gather information from the customer and a personalized recommendation is made at the end what sets Colette apart from other robo-advisor platforms is it here we're able to use the latest auto learning and natural language processing to really have a conversation around the customer's needs and circumstances and it's this that drives the product recommendation for each topic there are multiple levels of health information available depending on how much detail the customer needs a confident customer can skip the health information but a customer who needs more support can ask Colette for more details we've included different formats of explanation including embedding videos into the chat to cater to users with different communication styles if the customer has particularly complex needs or if Colette runs into difficulties the conversation can be delegated to a real advisor for the prototype we've integrated a connection to the live chat software live person there are three scenarios that could trigger a delegation if the customer has exhausted every level of health information and still doesn't understand a topic if the customer gives a series of bad rationales when asked to justify a decision or if Colette can't understand a customer's question when a delegation is triggered a summary of the conversations so far is transferred to the real advisor so they can quickly understand the context and they can resume the conversation in a live situation a bank adopting Colette would have a team of real mortgage advisors on hand to respond to these scenarios we will our course diversity and to technology Nina by nuance and the other was Watson by IBM these are two of the market leading technologies the name is analytics tool provides a way to log conversations and identify phrases that web recognized these can then be added to the corporate so that they are understood in future conversations we took two of the services the dialog service and the natural language classifier and what these two services do is one they'll take users input understand what the user means and have an appropriate response to that to the dialogue and the NLC will continually learn from user inputs and continually expand this knowledge base Watson's natural language classifier or NLC uses grammar rules and statistical algorithms to identify the meaning behind a customer's statement the NRC is trained by uploading sample texts each text is assigned a confidence level from 0 to 100 percent based on the strength of Watson's semantic classification threat matches can be confirmed and incorrect matches can be flagged helping to increase the accuracy of future classifications we can then define the confidence threshold above which a user input is accepted banks right now I expect you to deliver services 24/7 just like two girls fortify so using Colet as a way to deliver online advice solves a lot of challenges for them collect provides / benefits to customer you don't need to wait for an appointment in branch during applicants can progress the application from their living room and you don't need to worry about how you look I think we're at very exciting point in the evolution of artificial intelligence as it applies to banking we're really now a point of inflection where some of the on more leading-edge technologies have come of age and if you think about the the mass market that is the retail banking customer base and the way in which now the human-computer interface really has become a lot more natural it really allows you know that the artificial intelligence to be much more center stage in how banks deal with customers most across the world are only in real-time mortgage solutions banks are also under pressure to cut costs and that is why mortgage advisor solutions like Collette are really here to stay and other way from the future you