Names:
Anders Dræge
Christer Tellefsen
Kathrine Kleppestø
Roald Heie
Country: Norway
Organization: Frende Forsikring
Norwegian insurance company with around 300 employees, which is small by Norwegian standards. Continuously growing and focused on using technology to account for increased growth.
Awards Categories:
- Best AI Acceleration Use Case
- Best Moonshot Use Case
- Best AI Democratization Program
- Best ROI Story
Business Challenge:
To make claim reporting simple for our customers, we have one common email address for reporting claims. But we have four different claim units (except for life insurance): Buildings, Vehicles, Legal, Pets/Travel/Content. Emails used to be read and interpreted before manually forwarding them to the correct claim unit. The volumes are high, and the work is quite boring, so this was a perfect case for automation.
Business Solution:
We have trained a Bert model on around 10,000 emails to predict the correct unit for incoming emails. The precision of the model is high, and we manage to correctly predict claim units for 85-90% of all incoming emails. Some emails (like auto-replies or spam) are not supposed to be forwarded to any units, so the automation degree is even higher.
We have uploaded the trained Bert model together with the tokenizer into Dataiku. We have also used the Dataiku library to make functions to convert email attachments like PDF files, Word files, or images of text into digital text that is merged with the email text itself. All this is put into an API, which is made by the Dataiku API designer. The digital robot, which is actually performing the forwarding of emails, sends emails and attachments into the API, which returns the email destination together with a probability.
We are also exploiting the easy connection between Dataiku and Datawarehouse to see if we can identify the email writer as one of our customers and also the claim number, if there is any. If the customer is identified, we can find all open claims for the given customer. If there is a contradiction between open claims and the predicted claim unit, the email will be treated manually. Also, predictions with a lower probability than 0.7 will be treated manually.
But we are using the concept “Human in the loop” to improve the models over time: All emails that are sent manually to the claim units are stored by the robot. These are used to train the Bert model on selected emails that are particularly challenging. This has already improved the model, and we are expecting further improvement over time. All API responses are logged and available in a dashboard for relevant people.
This highly collaborative work which includes a variety of technologies and processes was performed by one person in the Team AI/ML and one person in the Team RPA (Robotic Process Automation). This is only possible because of the robust and streamlined platforms Dataiku and Blue Prism provide.
Day-to-day Change:
Our company is continuously growing, and we have stated in our strategy that the growth should be handled by increased automation and digitization. By combining AI and robotic processes, we have opened up new areas for automation and have played an instrumental role in reaching our long-term goals.
Business Area: Communication/Strategy/Competitive Intelligence
Use Case Stage: In Production
Value Generated:
This highly automated process has been estimated to save 56, 97, and 74 work hours in February, March, and April 2023, respectively. This frees up work hours for more complicated tasks that are less suited for automation. The process is scalable and puts Frende in a good position for growth, where we can handle larger volumes of emails in fewer work hours. We will also avoid risking that emails are forgotten or not handled within a reasonable amount of time, so the process is likely to have a positive impact on the response time and the total customer experience.
This project also served as a pilot for using AI with Dataiku together with a robot. Dataiku is well-designed for reusing and modifying existing projects, so we have implemented this technology in other areas as well. In total, we have seven Dataiku APIs in production that the robot uses for automating tasks and enriching the data with customer information.
Value Brought by Dataiku:
The Dataiku platform has allowed a small team to work very efficiently and produce highly advanced results in a relatively short time. Monitoring and communication of the automation and value creation are very simple through dashboards. The transparent and easily explainable workflow and results make it easier to get acceptance and understanding from both leaders and workers that are influenced by automation.
Value Type:
- Improve customer/employee satisfaction
- Reduce cost
- Save time
