Abstract The commercial loan origination process at the National Bank of Canada (NBC) historically faced challenges typical of complex, multi-actor financial services, including limited end-to-end visibility, significant process fragmentation, undocumented rework loops, and substantial manual effort. To address these inefficiencies, NBC partnered with Levio to implement a three-phase Process Intelligence approach. A human-in-the-loop prototype named Felix was developed to validate the interaction design, demonstrating the potential to reduce manual task time by an estimated 70% while maintaining strict regulatory compliance and audit traceability. This methodology offers a replicable framework for integrating AI into regulated, knowledge-intensive business processes.

Sophie Lacroix -  Senior Business Analyst - National Bank of Canada

Sophie Lacroix is a Senior Business Analyst at the National Bank of Canada, where she has built a career spanning over 20 years within the Commercial Banking division. She specializes in business analysis, business intelligence, and Agile methodologies, with hands-on expertise in credit analysis workflows, financial operations, and process optimization in regulated banking environments. She also served as a Product Owner from 2022 to 2023, bridging business strategy and technology delivery. Sophie holds a Bachelor's degree in Management Information Systems from the Université de Sherbrooke and is fluent in both English and French.

Frederic Brosseau - Partner, Process Intelligence Lead - Levio

Frédéric Brosseau is Partner and Process Intelligence Lead at Levio, a Canadian consulting firm. With over 25 years of experience in executive positions in technology and professional services, he specializes in process mining, intelligent automation, and digital transformation. He holds an MBA in Strategy from McGill University and a Bachelor of Business Administration from HEC Montréal. He has led process intelligence engagements across financial services, insurance, and public sector organizations

Abstract Healthcare pathways are notoriously hard to see: patients move across referral, imaging, pathology, and treatment steps that no single system links, and captures end to end. This talk describes how a regional cancer program applied process mining to make the real cancer journey visible — from suspicion to definitive treatment — and to quantify integrated wait times, turnaround, rework, and the points where bottlenecks emerge. Powered by a University of Ottawa co-op studentship, the program currently focuses on four parallel cancer pathway projects — breast, lung, and lymphoma assessment to definitive treatment pathways, and SBRT spine treatment planning — using one repeatable methodology aligned to the PM2 framework. We walk through how event logs were engineered from operational source systems, how variants and bottlenecks were validated with subject-matter experts, and how privacy was governed first through QI exemption, pseudonymization, and aggregate reporting. Beyond individual findings, the initiative produced reusable event dictionaries, project templates, and internal capacity in process mining. We close with candid lessons on standardizing method, protecting build time, and sharing results as knowledge exchange with the broader process mining community.

Julie Renaud, MRT(T), MBA — Director of Cancer Care at The Ottawa Hospital (TOH) and OH (CCO) Champlain Regional Cancer Program Director, and a Clinician Investigator with the Ottawa Hospital Research Institute (OHRI). She sponsors and leads the cancer pathway process-intelligence program, bringing clinical and operational leadership to the application of process mining in oncology.

Juan Ramirez-Acuña — University of Ottawa Master's student in Digital Transformation and Innovation (DTI), Faculty of Engineering. As a co-op studentship analyst  at The Ottawa Hospital, he focuses on event-log engineering and process analysis across the cancer pathway projects. Juan holds a Bachelor of Science in Computer Software Engineering from the Universidad Autónoma de Bucaramanga, Columbia.

Chang Liu — University of Ottawa Master's student in Digital Transformation and Innovation (DTI), Faculty of Engineering. Also a co-op studentship analyst at The Ottawa Hospital, contributing to data preparation, pathway mapping, and SME validation across the program. Chang holds a Bachelor's degree in Auditing from Nanjing Audit University, China, and a Master's degree in Applied Economics from the University of Michigan, USA.

Abstract Over the past two decades, several methodologies for carrying out Process Mining (PM) projects have been proposed. However, in sufficiently large organizations—where PM projects are often carried out by a Centre of Excellence—we have found that additional considerations come into play. These are not addressed in general-purpose PM methodologies, but are nevertheless pivotal in increasing the odds of a successful PM project. To address these considerations, we introduce the Process Mining for Large Organizations (PMLO) methodology, accompanied by artefacts such as questionnaires, based on our experiences in the public sector. PMLO iteratively covers six steps: 1) process qualification & selection, 2) project planning, 3) technical understanding, 4) data extraction & processing, 5) process mining, analysis, & evaluation, and 6) followup. We describe these steps and their sub-steps, and use case studies from large government departments to motivate and illustrate their usefulness. Early results suggest that PMLO provides practical benefits that cover early process qualification through questionnaires and interviews, stronger engagement with process owners, and support for continuous PM and post-improvement assessment. 

Jacques Trottier is the Process Automation Program Lead at Export Development Canada (EDC). He brings over 15 years of combined experience in digital transformation, IT, and business automation and championed the use of process mining in the Government of Canada.

Before joining EDC, Mr. Trottier was a Senior Advisor in Intelligent Automation at Health Canada and Manager of Intelligent Automation & Process Mining at Transport Canada, where he helped establish an automation centre of expertise delivering robotic process automation, workflow automation, AI, and process mining services. His team built an AI platform to screen air cargo for suspicious goods, transformed personnel security screening through workflow automation, and automated processing of zero-emission vehicle claims and other back-office functions. Mr. Trottier holds a BA and Master of Digital Transformation from the University of Ottawa.

Abstract In Predictive Process Monitoring, activity suffix prediction predicts the remaining sequence of activities given a prefix. As a multi-step sequence prediction task with many plausible suffixes, decoding is critical to predictive accuracy. However, the performance of existing decoding strategies is highly dataset-dependent, and no method is consistently reliable. A key limitation is a recurring mismatch between how suffixes are selected during decoding and how prediction quality is evaluated: current approaches typically rely on likelihood to guide decoding, without explicitly accounting for sequence-level prediction error. In this work, we characterize this mismatch and propose BRIDGE, a Bayes-risk-informed decoding framework that samples candidate suffixes and ranks them by expected quality under the evaluation metric, thereby aligning decoding more closely with the task objective. As a post-training inference-time method, it can be applied across predictive models without retraining. We evaluate BRIDGE on eight real-life event logs against multiple decoding strategies and established suffix prediction models. BRIDGE is consistently competitive and often best-performing, with particularly strong gains on logs with long traces and for simple statistical models.

Xiaomeng He pursues her PhD at KU Leuven, Belgium, where her research focuses on predictive process monitoring. She holds an MSc in Information Management, also from KU Leuven, a well as a Bachelor's degree, in Law from Peking University, China. She is currently Visiting Student Researcher at York University.

Abstract The Use Case Map (UCM) process notation, part of the ITU-T Z.151 User Requirements Notation standard, is used for requirements modelling and analysis, and is supported by the jUCMNav plug-in for the Eclipse environment. The PM4Py library (free!) was recently extended to support process mining  UCM models from event log (PM4Py-UCM, paper). This presentation will cover some of the benefits of UCM modelling over BPMN, and highlight many of the recent additions to PM4Py-UCM, including: UCM and BPMN visualization, binding of activities to resources with visualization, hierarchical decomposition of complex process models into sub-processes, concurrency-aware variants, export to jUCMNav with executable tests at the model level, decision mining from log attributes with condition minimization, activity/branch statistic computation, attribute-based process families, process metrics comparison, and interactive HTML export. A Streamlit-based Web interface is also available as a demo.

Daniel Amyot is Professor at the Faculty of Engineering, University of Ottawa, which he joined in 2002 after a few years as Senior Researcher at Mitel Networks. From 2002 to 2013, he led the development of the User Requirements Notation (URN) standard at the International Telecommunication Union (ITU-T). For nearly three decades, he also led the development of requirements modelling and analysis tools, including jUCMNav and, more recently, tools related to the Symboleo specification language. His research interests span many domains, including Software and Requirements Engineering, Business Process Mining and Modelling, Healthcare Informatics, Legal and Regulatory Compliance, and Smart Contracts. He holds a Ph.D. and an M.Sc. from the University of Ottawa (2001 and 1994), and a B.Sc. from Laval University (1992). He serves on the editorial boards of the Requirements Engineering (REJ) and Software and Systems Modeling (SoSyM) journals.

Beyond Process Discovery: Loan Origination Insights and he Human Side of Adoption 

Abstract As organizations continue to invest in process mining, success is measured not only by the insights generated but by the actions and adoption that follow. In this session, Hussam Gaoro shares lessons learned from a recent Loan Origination Process Mining initiative conducted at BDC using Celonis. Through this real world proof of value, the team explored process complexity, uncovered improvement opportunities, challenged existing assumptions, and developed a deeper understanding of how work actually flows across the loan origination journey. Beyond the findings themselves, this session explores the human side of transformation, including stakeholder engagement, the psychology of adoption, and practical strategies for turning process insights into meaningful business action. Attendees will leave with real world lessons, outcomes, and approaches for building momentum and long term value from their process mining initiatives.

Hussam Gaoro leads the Intelligent Automation Practice at BDC, driving initiatives in intelligent automation, process mining, and AI-enabled transformation. With extensive experience in enterprise automation, CRM transformation, and operational excellence, Hussam helps organizations build scalable foundations for successful AI adoption. A frequent speaker on process optimization, emerging technologies, and the human factors behind technology success, Hussam brings a practical perspective on achieving sustainable transformation. 

Abstract  Process mining tells us what happened. What teams actually keep asking is what is about to happen - which in-flight case will stall, breach its deadline, or be abandoned while there is still time to act. Answering that has normally needed a data scientist and a hand-built model, putting it out of reach of the analysts who know the process best. This talk walks end to end through building an early, actionable outcome prediction with no hand-written model, using Automatic Machine Learning, on a digital onboarding funnel where most applicants never finish. Then it turns to why most of these projects should not be trusted: prediction on event logs leaks, features quietly encode the answer, and the model looks brilliant. The takeaway is a practical recipe, and the discipline to know when the number in front of you is a lie.

Soren Frederiksen has more than 30 years’ experience in the fields of software development, artificial intelligence, and machine learning. He was Founder and CTO of iView Systems, a global award-winning software solutions company in the security and surveillance marketplace, which had a successful exit when it was acquire by Omnigo Software in 2017.

Abstract This research examines practitioners’ approaches to selecting automation opportunities, identifying technical feasibility with business impact, and measuring the automation impact on the improved business processes. The study’s qualitative research design involves semi-structured interviews with fifteen industry experts in Canada, including process analysts, automation leads, consultants, and technology specialists.

Najah El-Gharib is a PhD candidate in the Digital Transformation and Innovation program at the University of Ottawa. Her doctoral research explores the intersection of process mining and automation, with a focus on identifying, prioritizing, and enabling automation opportunities that enhance organizational efficiency and adaptability. In 2023, Najah co-founded the First Canadian Process Mining Community, which she continues to co-lead. The community connects experts from academia and industry to advance knowledge sharing and collaboration in process mining across Canada. She is also a part-time professor at uOttawa teaching a process modelling, mining and automation course. Alongside her academic work, Najah is a consultant in data analysis and AI, helping organizations leverage advanced analytics to drive innovation and operational excellence. 

Abstract Process mining is often touted as a data-driven technology, being based on the "raw" data recorded by information systems. Typically, however, an in-depth domain understanding is also needed, both to guide the process mining effort as well as interpret and analyze the mined process models. This semantic gap is often filled via qualitative analysis: meetings and interviews with domain experts. In this talk, I address our attempts at filling the semantic gap, both by squeezing more from the event log and by helping the domain expert during knowledge elicitation. A first approach involves distilling a domain ontology that encodes high-level domain semantics based on object interactions. In a second, explicitly human-guided approach, we leverage domain experts' ability to interpret specific log samples to elicit high-level domain semantics.

William Van Woensel received his PhD in Applied Computer Science at the Vrije Universiteit Brussel in Belgium. He was a Research Associate at Dalhousie University, and a Visiting Researcher at the Universidad Politécnica de Valencia, Spain. Themes of particular interest are Knowledge Representation and Reasoning, Process Mining, and, to a lesser extent, Mobile Computing. William is the co-chair of the W3C Community Group on Notation3, a declarative semantic language to implement decision making in open web environments. In the field of Process Mining, it is his ambition to integrate Knowledge Graph technology to more effectively cope with incomplete logs and models, answer process-related questions, and check conformance with normative models.

Abstract Process mining enables the discovery of process models from event data. However, traditional process discovery approaches often ignore contextual information or process attributes that may strongly influence real-world process behavior. This PhD research proposes a novel contextualized process discovery approach that exploits key contextual information to improve process analysis. The main idea is to use the information to enrich or abstract the discovered process model. 

Zahra Ahmadi pursues her PhD in Economics and Business at KU Leuven, Belgium, where she focuses on abstraction and contextualization in process mining. She holds an MSc in Computer Science from Allameh Tabataba'i University, Iran, and a BSc in Computer Engineering from Shahid Beheshti University, Iran. She is currently Visiting Student Researcher at the University of Ottawa.

Abstract This presentation showcases how process mining is advancing digital construction by transforming event data into actionable process insights through descriptive and diagnostic process analytics. It demonstrates how organizations are leveraging this technology to assess, monitor, and improve inter-organizational business processes across the AEC/FM (Architecture, Engineering, Construction, and Facility Management) industry, establishing a foundation for future context-driven predictive analytics and agentic AI process orchestration. Through real-world case studies in construction management processes and BIM-based process mining for Virtual Design and Construction (VDC), attendees will gain practical insights into how process mining enables data-driven process monitoring and control by revealing process inefficiencies, supporting better decision-making, and advancing digital project delivery.

Araham Martinez Lagunas is a construction professional driving the digital transformation of the construction industry through innovative, state-of-the-art technologies and methodologies aimed at enhancing efficiency and automation in construction business operations throughout the entire life cycle of construction projects. Araham's current efforts focus on advancing process intelligence in construction projects by leveraging data-driven and AI-powered methods, integrating the latest ICT research developments with practical industry applications to address the persistent challenge of productivity stagnation in the construction sector.

Araham holds a bachelor’s degree in Civil Engineering and a master’s degree in Construction Management from Tecnológico de Monterrey (Mexico), as well as a PhD in Building Engineering from Concordia University (Canada). He has over 10 years of experience in the construction industry, working on residential, industrial, institutional, and infrastructure projects.

Abstract Clinical guidelines tell emergency departments what care to deliver, but hospitals rarely have a way to measure whether that care actually happened. Conformance checking should be the natural fit, except that it asks the wrong question: it checks whether activities occurred in a modelled sequential order, while a guideline suggests a range of order independent actions given the patient's condition at that moment. In this talk, I present a state-based approach to clinical conformance checking. I model Ontario's Quality-Based Procedures as CMMN cases in which each recommended task is gated by a condition over the patient's recorded state, and I replay real emergency-department event logs against that model. The key idea is a task lifecycle that can withdraw a recommendation: when a severity score crosses a threshold, a test becomes due; when it falls back, that test is no longer due. In the second part of the talk, I show what this reveals on a real dataset of care that was genuinely missed, care that was reasonably skipped, and hundreds of tests performed on patients who never needed them, a form of non-compliance that conventional conformance checking rewards rather than flags. 

Vahid Kamandlooie is a Master's student in Digital Transformation and Innovation (DTI) at the University of Ottawa. His research focuses on process mining and compliance to quality-based procedures in a healthcare context, especially in emergency departments. Vahid holds a Bachelor's Degree in Mechanical Engineering from Arak University, Iran, as well as a Master's Degree in Business Administration and Management from Imam Khomeini International University, Iran.