AI-Enhanced Business Process Automation: A Case Study in the Insurance Domain Using Object-Centric Process Mining

Shahrzad Khayatbashi, Viktor Sjölind, Anders Granåker, Amin Jalali · 2025

Read AI-Enhanced Business Process Automation: A Case Study in the Insurance Domain Using Object-Centric Process Mining by Shahrzad Khayatbashi, Viktor Sjölind, Anders Granåker, Amin Jalali (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-17295.txt.

Introduction

Shahrzad Khayatbashi and co-authors wrote a 2025 case study of AI automation inside an insurance firm. A large language model took over spotting claim parts, a manual step that had capped growth. The team then mined the process with an object centred method that tracks how cases, claims, and parts move together.

Output capacity rose sharply with the model in place. New process patterns also appeared, needing refinement. The study shows the mining method working on live data, with its strengths and limits on show. Gradual automation, where old and new ways run side by side, is the real setting such tools must handle.

Catalog record

CREATORShahrzad Khayatbashi, Viktor Sjölind, Anders Granåker, Amin Jalali
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-17295

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

AI-Enhanced Business Process Automation: A Case Study in the Insurance Domain Using Object-Centric Process Mining Shahrzad Khayatbashi, Viktor Sjölind, Anders Granåker, Amin Jalali Abstract Recent advancements in Artificial Intelligence (AI), particularly Large Language Models (LLMs), have enhanced organizations’ ability to reengineer business processes by automating knowledge-intensive tasks. This automation drives digital transformation, often through gradual transitions that improve process efficiency and effectiveness. To fully assess the impact of such automation, a data-driven analysis approach is needed — one that examines how traditional and AI-enhanced process variants coexist during this transition. Object-Centric Process Mining (OCPM) has emerged as a valuable method that enables such analysis, yet real-world case studies are still needed to demonstrate its applicability. This paper presents a case study from the insurance sector, where an LLM was deployed in production to automate the identification of claim parts, a task previously performed manually and identified as a bottleneck for scalability. To evaluate this transformation, we apply OCPM to assess the impact of AI-driven automation on process scalability. Our findings indicate that while LLMs significantly enhance operational capacity, they also introduce new process dynamics that require further refinement. This study also demonstrates the practical application of OCPM in a real-world setting, highlighting its advantages and limitations. 1 Introduction Artificial Intelligence (AI), particularly in the form of Large Language Models (LLMs) [ 5 ] , is transforming business processes by automating knowledge-intensive tasks and enhancing operational efficiency. For example, identifying claim parts in the insurance domain is a knowledge-intensive task that traditionally requires human expertise due to the wide variety of claim parts. As claim volumes increase, manual identification becomes a bottleneck, limiting process scalability. LLMs present a significant opportunity to automate this task by leveraging their ability to understand context, recognize entities, and classify textual information based on prior training on vast datasets. Specifically, LLMs can be fine-tuned or prompted to extract relevant parts and classify them accurately within diverse claim descriptions, reducing reliance on human input and supporting digital transformation in business process management. Organizations often transition from traditional (as-is) processes to AI-driven (to-be) workflows gradually, resulting in a phase where both process variants operate in parallel. This coexistence allows organizations to evaluate the impact of AI automation on their ongoing processes, but it also presents challenges, as both process variants can influence the process outcomes. To support this, a data-driven approach should analyze both parallel process variants, each representing distinct aspects of the process, thereby enabling a more comprehensive understanding of how automation alters process behavior [ 9 ] . Object-Centric Process Mining (OCPM) [ 22 ] has emerged as a promising method for analyzing process transformations that enables the simultaneous analysis of multiple perspectives, making it particularly valuable for AI-driven process reengineering. Investigating side effects of process automation is important as indicated in literature [ 25 , 4 ] , where OCPM enable such investigation based on recorded process data. While previous research has demonstrated OCPM’s use in analyzing a single business process variant [ 2 , 21 , 15 ] , its ability to capture both as-is and to-be process variants simultaneously has yet to be investigated. To address this gap, this study presents a real-world case from the insurance sector, where the rising volume of claims necessitated AI-driven automation for claim part identification. The study follows the Business Process Reengineering (BPR) framework [ 18 ] to examine the transition from a manual as-is process to an AI-enhanced to-be workflow, supported by an LLM [ 5 ] for automation. Instead of fully automating claim part management, the LLM model generates leads for investigators, ensuring a balance between AI-driven efficiency and expert oversight and having humans in the loop. The AI implementation adheres to the CRoss-Industry Standard Process for Data Mining (CRISP-DM) methodology [ 7 ] , providing a structured approach to data understanding, modeling, and deployment. To evaluate the impact of AI-driven automation, this study applies OCPM to assess process scalability. The evaluation follows OCPM 2 [ 17 ] , an extension of the PM 2 process mining methodology [ 24 ] , enabling a systematic examination of process evolution. By applying OCPM, this study provides empirical insights into how AI-driven process reengineering enhances scalability while also assessing OCPM’s strengths and limitations in analyzing real-world process transformations. The remainder of this paper is structured as follows: Section 2 details the methodology, outlining the application of BPR, CRISP-DM, and OCPM for process evaluation. Section 3 presents the results, examining the impact of AI-driven automation on scalability and process dynamics while evaluating OCPM’s advantages and limitations. Finally, Section 4 concludes the paper. 2 Methodology Figure 1 : The followed research methodology based on BPR [ 18 ] , CRISP-DM [ 7 ] and OCPM 2 [ 17 ] , an extended version of PM 2 [ 24 ] . Figure 1 illustrates the Business Process Reengineering (BPR) framework [ 18 ] consisting of 6 phases that are followed in this study. In the understanding phase , the objectives of reengineering were defined, emphasizing the role of AI-driven automation in enhancing the scalability of the claim part identification process. This phase involved identifying scalability limitations and key challenges by analyzing the as-is process. Stakeholder interviews and process documentation were conducted to understand manual claim assessments, their associated bottlenecks, and the feasibility of AI-driven automation. Following this, the initiation phase defined the project vision, scope, and requirements. The primary goal was to develop an AI-driven approach for automating the identification of claim parts requiring further investigation. Additionally, data collection strategies were outlined to ensure necessary historical claim data availability for AI model training. Next, the model development phase (referred to as Programming in BPR) was carried out by the technical team, following the CRISP-DM methodology [ 7 ] to ensure a structured approach in developing a customized AI model, where its tasks are explained below: • Business Understanding , where the objectives and constraints of AI integration within the claims handling process were defined. Security approval was obtained to ensure compliance with security and privacy regulations. Discussions with domain experts helped establish requirements for claim part identification, ensuring alignment with operational goals. The primary business objective was to automate knowledge-intensive tasks while maintaining acceptable performance and enhancing scalability. • Data Understanding , where data sources, data quality issues, and relevant features were identified. The analysis focused primarily on claim descriptions and notes, representing the unstructured data sources. Additionally, exploratory data analysis (EDA) [ 6 ] was conducted to gain deeper insights. • Data Preparation , a labeled dataset was created by asking claim part investigators to label free-text claim descriptions and notes. This unstructured data served as the ground truth for evaluating the quality of the developed AI solution. • Modeling , where we leveraged the analytical capabilities of Large Language Models (LLMs) to identify claim parts. Various approaches were evaluated to ensure a consistent output format from GPT models [ 1 ] , with the Structured Output feature in OpenAI APIs found to be the most effective [ 20 ] . The prompt engineering process was a collaboration between business domain experts and data scientists. • Evaluation , where the LLM model was assessed using two approaches. First, quantitative

Read the complete text