New decision engine for EKF: How did it go?
Financing has become easier, but EKF’s new decision engine has resulted more assessments for the underwriter, says Kenneth Plum, Senior Vice President at Denmark’s Export Credit Agency.
One of the must-win battles in corporate strategy for EKF, Denmark’s Export Credit Agency, in 2018 was to strengthen our SME division. To do that, EKF embarked on another digital journey resulting from that strategic initiative.
The goals were to start an agile business development, increase credit quality, reduce the number of loss-making guarantees, make more time for the underwriter to handle demanding cases, improve the user experiences (for Danish banks), and secure future volumes of working capital guarantees.
Small group
EKF had a small group of people working in a hub a few kilometres from its headquarters in Nordhavn, Copenhagen. The group consisted of both underwriters, Credit/IT people, and the administration working with a consultancy company to learn to work ‘agile’ and to learn how to work out new processes, coding, testing, and so forth.
The solution consists of three elements: A front to the banks applying for working capital guarantees, a software robot, and a decision engine (Excel spreadsheet).
The project lasted for a little over six months and was concluded in 2019 as a fully functional robotic solution/minimum viable product (skateboard) that we can build (a car) on over time.
Took longer than expected
As it turned out, the project took longer than expected, and even with the perspective of building more features and increasing the elements of the goals, we had a huge setback in Q4 2019 with a breakdown of the front system connecting to the banks and later people close to the project leaving EKF. This meant that we haven’t come as far as we had hoped. COVID-19 also slowed down the development of ‘Decision Engine Version 2’ and a new customer portal due to changes in the prioritisation of resources.
However, the decision engine has helped change a guarantee from taking several hours to make to as little as 15 minutes.
The way it primarily helps the underwriter is that a lot of checking and information gathering is done by the robot. So instead of a long word document, all the underwriters have to check is the things listed in ‘yellow’ (a Yes or No to different parameters in our credit policies), and comment on those elements as all the ‘green’ pointers have been approved according to our credit policies.
That said, it’s only at this point that we can start to see the difference between now and then as a default normally only occurs two to three years after we have issued a working capital guarantee. The Danish government has provided different COVID-19 packages to Danish companies which have changed the normal credit/lifespan for private companies and those are very difficult to build into our solution, which is why it can be hard to see the real impact of the solution we have made.
What we learned
There have been a lot of benefits to this project – the reduction in time taken for underwriters being the first.
Also, the automated data collection that includes auditors’ notes, board overview, and additional data from the bank, including whether the company is risk-marked (OIK, or Objective Indication of Credit Deterioration), acceptance of GDPR, interest rates and other public data, etc. have given us a broader insight into the companies and in time will help us to sharpen credit decisions even more.
The project itself and the new way of working with robotics, data and the whole way of working agile have also given us ideas and courage to work with other automated solutions. For instance, when it comes to onboarding of new customers, handling ESG (external systems), and taking on changes in our systems as they are currently for more automated data sharing so that underwriters can put in data in one system (CRM) and it can generate reports to other branches so that the underwriters don’t have to fill out a lot of schedules when the data is already there.
Challenges we still face
The introduction of the decision engine was a tightening of the credit ratings, among other things, the requirement for the bank to disclose risk markers on the company (OIK). This tightening has meant that more cases must be dealt with by the underwriter.
The effect of the decision engine on EKF's losses is still difficult to assess because, as mentioned, a guarantee – based on EKF's current claims cases – a default first occurs on average two to three years after issuance.
Resources are an issue. It is an investment, no doubt about that, and since we started from the ground, we had to learn everything from there and slowly build up regarding production and internal setup to continuously focus on developing the solution and keeping it working on a day-to-day basis even when we started to put in more parameters such as changing suppliers.
Great impact
All in all, this has positively impacted the work we do at EKF and the mindset of how we want to attack challenges in the future. We are still a small group of people coming up with ideas and looking forward to implementing new features and making them more robust even though there is no further digital journey planned right now.