Implementing RPA bots at an insurance broker

Renaat Colpaert is Project Consultant at TriFinance ‘Financial Institutions’. When he started his mission, the client was starting the construction of their first bots. A traditional cycle of creating an RPA bot, consists of roughly the following steps:

• The identification of the processes to be automated
• The description of these processes
• The implementation of the bot
• Testing the bot
• The business acceptance of the bot

A first process had been selected and described, and was now being implemented. The implementation of such a bot consists of course of the installation of the software, but also and most importantly of the testing of the new bot. This first bot was meant to automate a simple but time consuming task: dispatching emails from clients to the correct team. The bot would search for the name of the client, find the corresponding file and forward it to the correct person to handle the request based on keywords found in the message.

The development of the bot was executed by an Indian partner. This partner based itself on the description of the selected process to make a first version of the bot. Testing the bot would then show whether the whole process was covered and whether the process was executed correctly. The bot was thus fine-tuned through several iterations of testing and readjusting the software. These iterations were sometimes a cumbersome but also a very gratifying experience. Testing an RPA bot indeed necessitates to be looking at a computer screen, watching the bot performing the tasks formerly performed by a colleague and checking whether the tasks are performed correctly. The gratifying part however is that the result is very tangible.

During these iterations, good communication skills are very important. One the one hand, Renaat had to be very precise in his communication towards the implementation partner to make sure they would understand exactly what was being expected. This way, he was ensuring a more efficient and effective construction of the bot. On the other hand, he first had to communicate with his colleagues at the client to get all the needed information concerning the process. A very thorough understanding of the processes was needed to be able to translate this towards the implementation partner. Being able to communicate in an efficient manner with both parties was a critical success factor in this project.

After this first successful implementation, a second set of processes was selected. This time, some more complex processes were chosen as well. One of the processes consisted of the analysis of hospital bills. A hospital bill can be quite complex as it consists of both refundable and non-refundable items, depending on the insurance policy of the insured person. The bill contains for example both the replacement of a hip and the food and drinks that were served during the hospital stay. Checking each line of the hospital bill versus the insurance policy is required to determine the amount to be refunded to the client. The bot was designed not only to analyze the bill and find the contract, but also to match both, and even to make the refund to the client's bank account.

While it sounds quite simple, the use of RPA makes it possible to automate complicated tasks as well in a relatively short time frame. While the implementation of a new application might take several months or even years, a bot can be implemented in only a few weeks. RPA is however not suited for the performance of all tasks. It is typically suited for repetitive tasks that have a standardized structure, and that will keep on being performed. In the case of an insurance broker, claims will come in on a daily basis and in great volumes through a web form for example. The handling of the most common claims is a task well suited for RPA.

The five main criteria to determine if a process is robotizable are as follows:

• The process should be rule-based. IF-THEN without interpretation.
• The tasks must be repetitive.
• The process should be stable & mature. Non-mature processes should first be standardized and streamlined.
• Data must be clean & structured: garbage-in garbage-out also applies for robotization.
• Activities should be time-consuming or high-volume transactions.

Some typical processes that meet these criteria are the following:

• The invoicing process: scan and register the invoice, release the payment, and notify the client
• Reporting processes: collecting, consolidating and reconciling data from different sources and presenting the data in a uniform format or dashboard
• Payroll activities: automate administrative tasks related to sick leave certificates, travel and expense policies, and the onboarding and off-boarding of employees
• Fraud detection: a lot of compliance checks on customer and transaction master data can be automated

Also to take into account when implementing RPA, is the application used for the task. RPA reproduces human actions which also means that it will for example steer the mouse to click on a certain button, copy information from a certain file, and fill out a certain field. If the layout of the application changes, and a button is not in the same place anymore or a field has been exchanged with another, the bot will fail. In Renaat’s project, the client had its own in-house application. This has as an advantage that the client can manage and foresee these kinds of changes if necessary.

This project is a fitting example of the TriFinance service offering around RPA:

Overall, we can support our clients in transforming their labor force into a hybrid, more digital labor force consisting of humans and robots by implementing RPA technology.

Furthermore, TriFinance can support you with RPA development. Combining RPA, programming, data quality, application development and the use of APIs, we deliver automation that can handle any complex environment while ensuring high manageability.

Depending on the needs, TriFinance will be more or less involved in the development of the RPA processes. In the case of Renaat’s project, the client continued its RPA project independently and has now automated more than 10 processes. Important to mention as well, is that in this mission, the people that performed the now automated tasks have all been given new and more challenging tasks thanks to the freed up time. The automation of processes thus doesn’t automatically imply a staff reduction.