In recent decades, Process Mining has played a key role in improving organizational processes. Today's growing organizational complexity, with collaborative processes involving multiple actors and interactions, demands new perspectives for Business Process (BP) design, analysis, and improvement. At the same time, there is increasing interest in reducing the environmental impact of human activities, in which processes are naturally involved. BPs sustainability evaluation is key for organizations to be able to improve their operations and minimize its impact. Artificial intelligence (AI) have gained relevance in all areas including process mining, to help and enhanced BPs evaluation and analysis, taking advantage of the large amounts of available process execution data. This doctoral work main objective is to expand the scope of process mining and AI by integrating techniques, practices, and tools for the design, sustainability and improvement of collaborative processes. The aim of this work is to reduce modeling efort, improve the execution and sustainability of collaborative processes, and reduce the gap between models and their actual execution. Validation of results will focus on controlled experiments and experts surveys, as well as case studies in a real context. The proposal will contribute to the field of PM for collaborative processes and sustainability analysis, helping organizations towards evidence-based BPs improvement.
Over the past decades, process mining [
Within collaborative processes [
In recent years, artificial intelligence (AI) has permeated all areas, including process mining, enabling
the integration of new approaches for design, automation, and data analysis, such as machine learning
(ML) for predicting future events based on historical event logs, or content generation using large
language models (LLMs), trained on vast amounts of data, to assist with diverse tasks such as process
modeling or execution trace analysis [
CEUR
ceur-ws.org
In this context, this doctoral thesis seeks to extend process mining with artificial intelligence and a sustainability focus, integrating new techniques, practices, methodologies, and tools to design and improve collaborative processes.The integration of diferent AI techniques, such as process data generation and analysis of traces through generative models, pattern recognition, or agent-based modelling supported by large language models (LLMs) will enhance process mining techniques providing new alternatives and results for sustainability analysis and process design and improvement. This work aims to ease modeling, enhance execution and sustainability in collaborative processes, and narrow the gap between models and execution.
The remainder of this document is organized as follows: Section 2 presents the motivation and research questions. Section 3 reviews related work. Section 4 describes the methodology, and Section 5 outlines the proposal. Finally, Section 6 presents the conclusions.
Process mining provides analytical tools that can be adapted to the proposed context, even in collaborative processes, and within traditional event logs or object-centric event logs. Few proposals address the automation and improvement of BPs with a sustainability focus; a key enabler is collecting and integrating sustainability data into their execution for further analysis and tool support.
Also, the current state of several artificial intelligence applications ofers a new set of readily available
methods and tools, such as ML, LLMs and Agentic-based approaches (c.f. Section 3, which the integration
within process mining techniques and perspectives we aim to asses and provide within the approach.
Based on the previous statements, the research question is defined as follows: How can process mining
be extended with artificial intelligence methods to capture, analyze, and improve sustainability aspects
in collaborative processes?. From the general research question, specific research questions are defined:
• RQ1. What process mining-based techniques, practices, methodologies and tools could be useful
for the design, sustainability analysis and improvement of collaborative business processes?
• RQ2. What process mining approaches exists to deal with business process sustainability and
which artificial intelligence methods are integrated, in particular for collaborative processes, and
what limitations they present?. The answer to this research question, which will be obtained by a
systematic literature review [
The application of process mining techniques (discovery, conformance, enhancement, prediction) on
business process execution data [
On the other hand, sustainability-oriented processes and approaches such as Green BPM have gained
relevance. In [
The evaluation of ICTs in [23] aims to manage energy resources eficiently while controlling the impact of the performed activities, and [24] analyzes how supporting software, such as BPMS platforms, artificial intelligence algorithms [25], and process discovery techniques [26], influences energy consumption. The notion of “Green Data Science” (GDS) was introduced in [27] due to the potential “pollution” that data science itself can cause.
Advances in artificial intelligence have enabled the integration of new automation and data analysis
approaches, such as machine learning (ML) for predicting future events [28, 29], or the creation of new
content through large language models (LLMs) to assist in various tasks. In particular, LLMs have been
used to generate textual descriptions from process data and event logs, to derive process models from
textual descriptions, and to leverage textual descriptions of models and event logs for process mining
[
To guide the research, the principles of Design Science are followed, [31, 32, 33]. focusing on the creation
of artifacts such as: a framework and working methodology, diferent models (processes, ML, LLMs, and
prompt engineering), and algorithms for process mining, data recording and management, sustainability
measures, among others. Following this methodology, the research started with the problem definition
to identify hypotheses and research questions, followed with a state-of-the-art review on the selected
topics, based on the guidelines for systematic literature reviews and mapping studies proposed in
[
For the validation of the diferent results, the guidelines in [ 31, 34] will be followed, considering both the technical validation and the empirical validation of the proposed artifacts. For instance, controlled experiments and surveys will be carried out to validate the adequacy of the proposed solutions with domain experts (e.g., sustainability measures and analysis), to gather their opinions on the usefulness, applicability, and suitability of the proposals in real-world contexts, and with functional prototypes. Furthermore, case studies [34, 35] are planned within the context of real collaborative e-Government processes (AGESIC)1, which will allow assessing applicability as well as identifying opportunities for improvement and limitations.
The proposal integrates sustainability into the entire process life cycle, from design and implementation to execution, analysis, and improvement. The focus is on collaborative business processes supported by information systems, particularly BPMS, though it can also extend to other domains. Figure 1 provides an overview, illustrating a BPMS for BPs execution as example.
A key component is a systematic approach for registering sustainability-related data during process execution (i.e. parameters and measures values), drawing, for instance, on the measures described in [23, 19] with corresponding calculation formulae. Conceptual components and tools support this registration by capturing parameter values at execution time. They also distinguish between estimated and measured data, helping users interpret the resulting values appropriately. From process execution, event logs are generated, containing sustainability-related data registered directly during execution i.e. parameters values (and optionally also measures calculated values), as well as adding measures calculation after a log enrichment process, using as basis the registered data i.e. parameters values. 1https://www.gub.uy/agencia-gobierno-electronico-sociedad-informacion-conocimiento/ These options are described in subsection 5.1, and applies for traditional XES and object-centric event logs.
With the sustainability extended event logs, process mining is performed using traditional approaches,
complemented by the new possibilities ofered by artificial intelligence tools such as Large Language
Models (LLMs) [
Existing approaches that incorporate sustainability attributes into business processes do not clearly specify how such data can be systematically registered and derived. To fill this gap, this work proposes a systematic approach that integrates concrete sustainability measures based on [23, 19], whose values are either registered directly during execution or calculated in a post-processing stage using sustainability parameters registered during runtime. The focus is on the enviromental dimension, including categories such as energy, emission, material, water, waste and software. We extended them by adding the concrete formulae and parameters needed for calculation, and adding measures related to BPs execution in an information system, such as the language of implementation e.g. Java, Python, etc, the CPU, cloud, among others, which can be estimated based on existing reference data, such as CO2 emissions.
The proposal includes a method to register and calculate sustainability measures, an extension to the XES format to include sustainability data (and translation to object-centric event logs), specific sustainability measures and a measurement taxonomy that distinguish between estimated and actually measured values (e.g. with a phisical devise such as sensors) diferentiating approaches by their degree of empiricism, to help users in interpreting the results. Tool support is also provided for sustainibility data registration and microservices-based sustainability measures calculator.
Various artificial intelligence tools can be applied to process mining, as has traditionally been done with machine learning, and more recent studies are focusing on LLMs. Evidence, such as that presented in [37], shows satisfactory performance in a range of process mining tasks using large-scale commercial LLMs. However, it is worth noting that the environmental impact of LLMs can be considerable when deployed at a global scale [38]. For this reason, alternative approaches will be explored that leverage the capabilities of the technology available for process mining while, from a sustainability perspective, ensuring that the tools themselves do not become part of the problem. Through the concept of agents, multiple specialized process mining tools can be orchestrated in a coordinated manner across diferent stages of the process mining lifecycle to enhance their tasks. For example, the integration with Model Context Protocol (MCP) [39] servers to allow specific context for LLMs use will be evaluated.
The integration of data associated with processes and organizational information, particularly sustainability-related data, is key in the context of the proposed work. In previous work the integration of organizational data into process logs [40] was addressed, defining methodologies and tools that enable a joint process and organizational data analysis. This will be translated also into OCED event logs and analsys, leveraging the metamodel and implementation described in [41], with focus on collaborative processes and associated challenges e.g. data privacy and security across organizations, common goals for sustainability, heterogeneous infrastructure, among others. On the other hand, since OCED is still a relatively recent standard, there are few process logs available to support research tasks. In this regard, several strategies are being analyzed to generate synthetic event logs based on OCED such as Conditional Tabular GANs (CTGANs) [42]. Preliminary results indicate that it is possible to generate high-quality synthetic logs using these technologies, which will serve as a valuable input for experimentation with diferent datasets.
Based on the process mining with IA support analysis, integrating sustainability data and measures,
improvement recommendations will be derived to help organizations improve their processes, with focus
on collaborative scenarios. By adding the sustainability dimension to the Devil’s Quadrangle similarly
to [
This work addresses the challenges inherent in improving collaborative processes by incorporating an environmental sustainability perspective through process mining, supported by advances in artificial intelligence. These challenges are addressed through the definition of several artifacts which will be validated from diferent perspectives with controlled experiments and surveys, and real-world cases of collaborative processes within the Uruguayan government, employing data provided by AGESIC.
Suitable techniques, practices, methodologies, improvement recommendations and supporting tools will be integrated and defined for the design and redesign of sustainable processes, taking into account and extending/adapting existing proposals, as well as defining new ones to achieve the stated objectives. The results of this work will be of help to organizations in order to analyze and improve their collaborative BPs with a sustainability focus, minimizing their environmental impact.
Partially supported by projects: “Minería de procesos con automatización robótica e IA generativa para el diseño y sostenibilidad de procesos colaborativos hiperconectados” funded by Comisión Sectorial de Investigación Científica (CSIC), Universidad de la República, Proy I+D 2024, “22520240100504UD”, Uruguay and Academic Graduate program PEDECIBA Informática, Ministry of Education and Culture (MEC) and UdelaR, Uruguay.
During the preparation of this work, the author used Chat-GPT-5 in order to: Text Translation and Grammar and spelling check. After using these tool, the author reviewed and edited the content as needed and take full responsibility for the publication’s content.