Professional and holistic debt counseling requires competence, empathy and tact towards clients who require support in a very personal and difficult situation. Furthermore, from the debtor’s point of view, it is important to receive support and advice at short notice. The focus of the debt counseling firm Schulz & Partner Attorneys at Law is to allow little to no waiting time for its clients, and to represent them professionally vis-à-vis creditors and the courts.
The law firm has obtained support in order to be able to focus 100% on these important activities. Since the beginning of 2022, “digital employees” have taken over manual back-office activities, so that every member of the firm has more time to provide clients with individual advice and support. Software robots, incorporating artificial intelligence, consistently take over successive activities such as creating a file, determining the amount of the claim for each debtor, or sending settlement offers to creditors. In this way, employees of the law firm have more time for topics that no “robot” can or should take over – namely important and sensitive communication with clients, creditors and courts.
Project description – preparation
With the goal of relieving employees of manual back-office tasks, Schulz & Partner first analyzed the overall process – from the receipt of a debtor counseling contract to the final determination of the funding amount per creditor. The overall process was then divided into individual process steps and the automation potential was identified in these sub-processes.
In order for the firm to assess both the technology used and the project implementation partner “risk-free” in projects of a manageable size, individual tasks were first automated together as part of a pilot project. In this way, Schulz & Partner was able to convince itself of the result in productive operations and subsequently automate further processes. It quickly became apparent throughout the company that the implementation was precise, robust and stable, and that employees were noticeably relieved of their workload. On this basis, the degree of automation is being successively increased in view of the overall processes.
Framework parameters of the pilot project
The “Processing of incoming receivables letters” process was selected for the pilot project. When selecting this process, various evaluation criteria were included in the decision:
- A rule-based process with a high number of cases and stable systems
- High case numbers are processed across systems
- A high level of manual effort
The evaluation criteria for a successful pilot project were met by the process. The goal of the pilot project was to keep the initial effort low on both sides in the first step in order to achieve results quickly. To ensure this, milestones were defined during the course of the project. When a milestone was reached, the result was evaluated together with the customer and a decision was made whether to continue or end the project. Furthermore, within the pilot project not all demand letters were read out. Instead, the quantitative “top demand letters” were automated first. For this purpose, the data from the demand letter relevant for process automation was defined. The cover letters were then analyzed and their readability evaluated. The result? Due to the high scanning quality of the demand letters on the part of the customer, the letters were suitable for automation.
Automation of the “Top Five” demand letters
As part of the pilot project, the “Top Five” demand letters were selected for automation. Different automation technologies were used from a “digital toolbox”:
- Text Recognition (OCR) software is used to scan the demand letters.
- Document Understanding (DU) is used to extract the relevant data from the demand letters, which are structured differently for each creditor letter.
DU is AI-powered software that enables data extraction and processing from documents. The AI model can be trained independently by the client via a web-based user interface, which has the advantage of increasing the confidence of the AI model in a resource-efficient manner. Due to the large number of differently structured demand letters and the amount of data to be read out, implementation without AI is not possible. The data read out by the DU is then transferred to a software robot, which uses Robotic Process Automation (RPA) to transfer the data to the underlying systems. RPA enables the automation of manual process steps in the user interface of applications where data is collected and applications are operated like a user. In doing so, RPA works in a minimally invasive manner and supports manual, repetitive activities across systems. As a result, the use of RPA transformed a previously 100% manual process into a 93% automated process.
Within the process, the robot may need assistance in validating the data or deciding on further processing. The robot uses a so called “Action Center” to communicate with the clerks through a web-based interface and then continues working after receiving the missing information from human intervention.
Automation of the overall process
Considering the overall process described in the preparation, after the successful completion of the pilot project the aim was to expand the automation journey. Additional upstream and downstream processes were added to the receivables registration process:
- In the first step, the AI model was continuously extended and improved by the customer, so that with the help of human-machine interaction all claim letters are processed automatically.
- The upstream file creation, as well as the downstream processes to validate a settlement offer and the subsequent response to the settlement, were automated. For each of the processes, DU is used to read the data.
- A downstream RPA process is used to automate the customer systems. Again, humans work hand-in-hand with the robot via the Action Center.
- In the “Create debtor file” process, creditor data is read from a PDF and then the associated debtor file is opened and expanded to include the creditor data that was read in.
- The task is then set to be completed in the workflow management tool used.
- In the downstream process Evaluation Settlement, the settlements created by the customer are validated and documented using the master data supplemented by the previous processes.
- After the customer has sent the settlement offer, the robot waits for the corresponding response from the creditor. Once a response has been received, the robot accepts the response and reads the predefined data from the response letter.
- Based on the letter, the AI model detects whether the creditor accepts, rejects, or makes a counteroffer to the settlement.
- If the offer is accepted or rejected, the response is mirrored in the customer systems. If it is a counteroffer, the terms are documented and communicated to the clerk.
As a result, the manual effort per debtor file has been reduced from 15 minutes to 2 minutes.
Due to the modular development of the robots, the processes and AI models can be constantly optimized. On the one hand, the continuous improvement increases the degree of automation and on the other hand, reduces human-machine interaction. The practice with automation has shown that further process optimizations and process ideas emerge after the completion of a successful pilot.
The value-driven approach using RPA has sharpened the focus on process automation at Schulz & Partner and demonstrated the possibilities for handing over manual activities from employees and staff to “digital staff”. With a focus on automation, the project team is constantly identifying further optimization approaches. In terms of technology, the “artificial intelligence” is constantly being trained and focused on core tasks. In addition, the project team has identified the potential to enable employees to hand over tasks to a “digital assistant” directly at the workplace, as needed and “at the push of a button”.
Automation with RPA is a journey. You have to start this journey with a first step, which can be exciting and promises to leverage further automation potential in the future.