Bankdata joins Jyske Bank in providing machine learning for 1.5 million customers
Bankdata is the leading Danish IT support organization that delivers IT solutions to 11 banks in Denmark, covering 1.5 million banking customers.
In January 2017, Bankdata purchased Intellix (AI) machine learning system to be able to easily capture and automate specialist banking expertise.
Documentation that Bankdata's 1.5 million customers are receiving consistent best practice advice is not only securing good customer service, it is also a necessity for legislation compliance.
Jyske Bank started using Intellix' machine learning system for its 1,500 banking advisors in 2002, and has been successful in doing so, consequently leading Bankdata to make a deal with Intellix.
Investment advisory is a complex system in which the advisor guides the customer on investment portfolios based on their financial situation, risk willingness and goals.
To value how much risk the customer is willing to tolerate is one of the most complex and difficult tasks for an investment advisor. It’s not just about cold figures. Income, expenses and risk willingness is not black and white, and a banking advisor does not simply pose the question "how much are you willing to risk losing on a yearly basis" in order to attain sufficient risk profiling. It involves both the ability to take a loss as well as how the customer would feel when the market declines. An investment advisor may view this tentatively while another customer would see it as a buying opportunity. Using machine learning from Intellix makes it easier to weed these complexities out by feeding the system with case examples from which it learns and provides answers.
»We do not know what our risk profile is even though we are asked directly«, says Peter Bruun, department manager for investment services at Bankdata.
»It is about testing the customer in relation to facts, feelings and consequences. It is a very demanding area for the investment advisor«, states Peter Bruun.
Bankdata uses Intellix' standard knowledge automation software that entails neural networks to crunch the input that the banking advisors collect from the customers. Intellix then automates how much the customer can afford to lose and how much he or she is willing to lose.
»For a banking advisor it means collecting knowledge assembled from all the banks’ best advisors and giving a unified second opinion to customers. We cannot get this precise efficiency in any other way«
Best-practice data
Bankdata delivers financial IT solutions and services to the following banks: Alm. Brand Bank, Djurslands Bank, Jyske Bank, Kreditbanken, Nordfyns Bank, Nordjyske Bank, Ringkjøbing Landbobank, Skjern Bank, Sparekassen Sjælland-Fyn, Sydbank og Østjydsk Bank, which together cover over 1.5 million customers.
The system will be implemented in October this year for nine of the eleven banks, while Sydbank will convert their existing solution as well to Intellix in the new year and Jyske Bank will have their existing Intellix solution hosted by Bankdata in 2018.
Example of how the Intellix system works
Intellix presents the software with a simplified case from the insurance industry; how to build a fully automated system that can delineate a client’s health status by using AI neural networks to capture an expert’s tacit knowledge. Link to the presentation here.
In the Intellix system an expert provides examples of people and their specific health status based on a number of factors. The examples provided are the most obvious examples of healthy and unhealthy people and these are the ones that are easy for the expert to provide from his active knowledge, also called explicit knowledge.
»Each and every case has been validated, and peer reviewed«, Peter Bruun explains. »We deliver quality based on the data provided by the most experienced advisors«.
Finds uncertainty
When the Intellix system is trained with specific and precise cases, it is able to identify areas where the model is weak. Thereafter, Intellix would provide examples of cases in which the system is uncertain thereby activating the expert’s passive knowledge, or also called tacit knowledge, for him/her to decide what the results should be or further information is needed. This tacit knowledge acquisition capability is unique to the Intellix software, which learns and unlearns in real-time and stops when the system is capable of making correct decisions. Intellix is also able to identify which factors have an impact on the results and to what degree. Such factors that bare no impact can be deleted, unless they need to be asked based on legal compliance factors.
The result is a machine learning system which uses sufficient data from several top banking advisors and delivers easily maintainable and documented investment advisory systems that are consistently better than the best individual banking advisor.
The system is devised to use conservative measures in order to provide less risky decisions in case of uncertainty.
»We would not recommend an investment that is specifically risky. The system is set up to lean towards more secure outcomes when the decision is unclear«, explains Peter Bruun.
Online self-service
The Intellix system is implemented as a support system for the banking advisor to be used in person-to-person meetings during which the advisor will ask the customer for information with which to feed the system.
There are plans thereafter to implement the same Intellix system as self-service for bank customers to use within the on-line banking portal.
»Here, there is a possibility to integrate the data we already have regarding the customer, subject to customer acceptance, of course«, says Peter Bruun. Even though the system comes with certain advice it ultimately will be the customer who makes the choice in what to invest. In such cases the system will be able to document that a customer was advised against for example a specifically risky investment.
With the utilization of Intellix software, Bankdata will ensure that banking advisors consistently use best and trustworthy practices - the best collective knowledge within the organization - when providing its customers with investment advice.
As we saw during the 2008 financial crisis, one of the worst-case scenarios upon investing is when there is a discrepancy between what the bank and the customer expects«, finishes Peter Bruun.
The article in Danish published on September 21, 2017 in Version2.dk can be viewed here.
