Marco Schreyer
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Marco Schreyer has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Miklos A. Vasarhelyi
Rutgers, The State University of New Jersey
3 shared publications - Timur Sattarov
Research Data and Service Centre, Deutsche Bundesbank, Frankfurt am Main, Germany
3 shared publications - Damian Borth
School of Computer Science, University of St.Gallen (HSG), St.Gallen, Switzerland
2 shared publications Kai-Uwe Marten
Ulm University
2 shared publications- Qing Huang
Marshall University
2 shared publications - Tassilo Lars FÖHr
Ulm University
2 shared publications
Latest Publications
- Open Access
- 11 September 2026
- Published by Elsevier in SSRN Electronic Journal
Auditing Artificial Intelligence: A Research Agenda for AI Assurance
Artificial intelligence (AI) is becoming embedded in organizational processes, raising a fundamental question for auditors: how can they provide credible independent assurance over systems whose risks and dependencies evolve across models, data, components, organizational processes, and deployment environments? As these systems become more capable, interconnected, and context-dependent, established assurance principles remain relevant but become more difficult to apply. This paper develops a research agenda for auditing AI. First, it examines how advances in AI capabilities, system architectures, and deployment settings change what must be assured, which risks matter, and where to draw engagement boundaries. Second, it synthesizes developments in policy, regulation, professional and standard-setting activities, and technical AI research to characterize the emerging AI assurance ecosystem. Third, building on the risk-based auditing process, it identifies seven interconnected research directions spanning engagement design, criteria, procedures, evidence, professional judgment, reporting, and the conditions needed to sustain credible assurance. Together, these contributions provide an organizing framework for research on a central challenge facing the auditing profession: how to develop and sustain credible assurance as AI systems and their organizational uses continue to evolve.
- 11 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAuditing Artificial Intelligence: A Research Agenda for AI Assurance
Artificial intelligence (AI) is becoming embedded in organizational processes, raising a fundamental question for auditors: how can they provide credible independent assurance over systems whose risks and dependencies evolve across models, data, components, organizational processes, and deployment environments? As these systems become more capable, interconnected, and context-dependent, established assurance principles remain relevant but become more difficult to apply. This paper develops a research agenda for auditing AI. First, it examines how advances in AI capabilities, system architectures, and deployment settings change what must be assured, which risks matter, and where to draw engagement boundaries. Second, it synthesizes developments in policy, regulation, professional and standard-setting activities, and technical AI research to characterize the emerging AI assurance ecosystem. Third, building on the risk-based auditing process, it identifies seven interconnected research directions spanning engagement design, criteria, procedures, evidence, professional judgment, reporting, and the conditions needed to sustain credible assurance. Together, these contributions provide an organizing framework for research on a central challenge facing the auditing profession: how to develop and sustain credible assurance as AI systems and their organizational uses continue to evolve.
- Open Access
- 5 August 2026
- Published by Cappelen Damm AS - Cappelen Damm Akademisk in Magma
- Vol. 29 (3)
Agentic Authority and the Future of Organisational Leadership
The rise of agentic artificial intelligence (AI) is reshaping organisational authority by introducing a new form of influence: agentic authority. Unlike traditional hierarchical authority grounded in a formal position or professional expertise, agentic authority derives legitimacy from data, predictive accuracy and system design. This article develops a conceptual framework for analysing how AI-mediated authority affects leadership practices and organisational power relations. It argues that leadership increasingly operates at the intersection of human judgement and agentic recommendation, creating new challenges concerning responsibility, legitimacy, accountability and trust. The article proposes that future leaders must act as ‘authority integrators’, mediating between data-driven prescriptions and normative organisational values. By examining how agentic authority interacts with traditional forms of managerial authority, the article advances leadership theory beyond person-centred models towards a hybrid understanding of governance in AI-enabled organisations.
- 5 August 2026
- Published by Cappelen Damm AS - Cappelen Damm Akademisk in Magma
- Vol. 29 (3)
Open AccessAgentic Authority and the Future of Organisational Leadership
The rise of agentic artificial intelligence (AI) is reshaping organisational authority by introducing a new form of influence: agentic authority. Unlike traditional hierarchical authority grounded in a formal position or professional expertise, agentic authority derives legitimacy from data, predictive accuracy and system design. This article develops a conceptual framework for analysing how AI-mediated authority affects leadership practices and organisational power relations. It argues that leadership increasingly operates at the intersection of human judgement and agentic recommendation, creating new challenges concerning responsibility, legitimacy, accountability and trust. The article proposes that future leaders must act as ‘authority integrators’, mediating between data-driven prescriptions and normative organisational values. By examining how agentic authority interacts with traditional forms of managerial authority, the article advances leadership theory beyond person-centred models towards a hybrid understanding of governance in AI-enabled organisations.
- 1 February 2026
- Published by American Accounting Association in AUDITING: A Journal of Practice & Theory
- p. 1-27
Connecting the Dots: Graph Neural Networks for Auditing Accounting Journal Entries
SUMMARY: Based on the double-entry bookkeeping mechanism, each transaction is recorded in at least two ledger accounts, with one debit and the other one credit. Journal entry data, in the context of accounting, contains a rich network of information that can be effectively translated into a graph. This study explores how to use graph neural networks to learn graph representations from journal entry data and to systematically understand the complex patterns and connections that exist in journal entries at the transaction level. The real-world application results demonstrate that the unsupervised graph neural network framework offers a promising methodology for detecting error and fraud in financial audits.1 Data Availability: Dataset A comes from the Ernst & Young Academic Resource Center (EYARC), https://www.eyarc.site/blog/peach-state-university-hotel-understanding-audit-analytics-case
- 1 February 2026
- Published by American Accounting Association in AUDITING: A Journal of Practice & Theory
- p. 1-27
Connecting the Dots: Graph Neural Networks for Auditing Accounting Journal Entries
SUMMARY: Based on the double-entry bookkeeping mechanism, each transaction is recorded in at least two ledger accounts, with one debit and the other one credit. Journal entry data, in the context of accounting, contains a rich network of information that can be effectively translated into a graph. This study explores how to use graph neural networks to learn graph representations from journal entry data and to systematically understand the complex patterns and connections that exist in journal entries at the transaction level. The real-world application results demonstrate that the unsupervised graph neural network framework offers a promising methodology for detecting error and fraud in financial audits.1 Data Availability: Dataset A comes from the Ernst & Young Academic Resource Center (EYARC), https://www.eyarc.site/blog/peach-state-university-hotel-understanding-audit-analytics-case
- Open Access
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 2478-2489
Diffusion-Scheduled Denoising Autoencoders for Anomaly Detection in Tabular Data
- 3 August 2025
- Published by Association for Computing Machinery (ACM)
- p. 2478-2489
Open AccessDiffusion-Scheduled Denoising Autoencoders for Anomaly Detection in Tabular Data
- Open Access
- 14 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 617-625
Imb-FinDiff: Conditional Diffusion Models for Class Imbalance Synthesis of Financial Tabular Data
Handling imbalanced datasets remains a critical challenge in financial machine-learning applications such as loan approval, credit scoring, and fraud detection. We present Imbalanced Financial Diffusion (Imb-FinDiff), a novel denoising diffusion framework designed to address class imbalance in financial tabular data. Our framework leverages embedding encodings for categorical and numerical attributes, effectively managing the complexities of mixed-type financial datasets. By incorporating a dual learning objective, (i) diffusion timestep noise and (ii) class label prediction, we synthesize minority class samples. Extensive experiments on diverse and real-world financial datasets demonstrate that Imb-FinDiff maintains the statistical properties of the original data while reducing bias caused by class imbalance. The minority class samples generated by Imb-FinDiff enhance the utility and fidelity of downstream machine learning classifiers.
- 14 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 617-625
Open AccessImb-FinDiff: Conditional Diffusion Models for Class Imbalance Synthesis of Financial Tabular Data
Handling imbalanced datasets remains a critical challenge in financial machine-learning applications such as loan approval, credit scoring, and fraud detection. We present Imbalanced Financial Diffusion (Imb-FinDiff), a novel denoising diffusion framework designed to address class imbalance in financial tabular data. Our framework leverages embedding encodings for categorical and numerical attributes, effectively managing the complexities of mixed-type financial datasets. By incorporating a dual learning objective, (i) diffusion timestep noise and (ii) class label prediction, we synthesize minority class samples. Extensive experiments on diverse and real-world financial datasets demonstrate that Imb-FinDiff maintains the statistical properties of the original data while reducing bias caused by class imbalance. The minority class samples generated by Imb-FinDiff enhance the utility and fidelity of downstream machine learning classifiers.
- Open Access
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Connecting the Dots: Graph Neural Networks for Auditing Accounting Journal Entries
Based on the double-entry bookkeeping mechanism, each transaction is recorded in at least two ledger accounts, with either debit or another credit. Journal entry data, in the context of accounting, contains a rich network of information that can be effectively translated into a graph. This study explores how to use graph neural networks to learn graph representations from journal entry data and to systematically understand the intricate patterns and connections inherent in journal entries at the transaction level. The real-world application results demonstrate that the unsupervised graph neural network framework of journal entry data offers a promising methodology for detecting fraud and error in auditing work.
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessConnecting the Dots: Graph Neural Networks for Auditing Accounting Journal Entries
Based on the double-entry bookkeeping mechanism, each transaction is recorded in at least two ledger accounts, with either debit or another credit. Journal entry data, in the context of accounting, contains a rich network of information that can be effectively translated into a graph. This study explores how to use graph neural networks to learn graph representations from journal entry data and to systematically understand the intricate patterns and connections inherent in journal entries at the transaction level. The real-world application results demonstrate that the unsupervised graph neural network framework of journal entry data offers a promising methodology for detecting fraud and error in auditing work.
- Open Access
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Deep Learning Augmented Risk-Based Auditing
Due to recent advances in Deep Learning (DL) such as GPT-4, the auditing and accounting profession should strongly consider its use in risk-based auditing. Although prior studies have already examined the use of DL in risk-based auditing, they lack general guiding principles to lower the auditor’s inhibition threshold when applying DL for audit procedures. We take the first step towards establishing such guiding principles by proposing a framework based on the CRISP-DM process model and AI Canvas for the implementation and utilization of DL in risk-based auditing. Therefore, we conducted focus group discussions as well as expert interviews with subject matter experts in DL as well as auditing to extract feature attributes necessary within each phase of our developed framework.
- 1 January 2024
- Published by Elsevier in SSRN Electronic Journal
Open AccessDeep Learning Augmented Risk-Based Auditing
Due to recent advances in Deep Learning (DL) such as GPT-4, the auditing and accounting profession should strongly consider its use in risk-based auditing. Although prior studies have already examined the use of DL in risk-based auditing, they lack general guiding principles to lower the auditor’s inhibition threshold when applying DL for audit procedures. We take the first step towards establishing such guiding principles by proposing a framework based on the CRISP-DM process model and AI Canvas for the implementation and utilization of DL in risk-based auditing. Therefore, we conducted focus group discussions as well as expert interviews with subject matter experts in DL as well as auditing to extract feature attributes necessary within each phase of our developed framework.
- Open Access
- 25 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 64-72
FinDiff: Diffusion Models for Financial Tabular Data Generation
The sharing of microdata, such as fund holdings and derivative instruments, by regulatory institutions presents a unique challenge due to strict data confidentiality and privacy regulations. These challenges often hinder the ability of both academics and practitioners to conduct collaborative research effectively. The emergence of generative models, particularly diffusion models, capable of synthesizing data mimicking the underlying distributions of real-world data presents a compelling solution. This work introduces Financial Tabular Diffusion (FinDiff), a diffusion model designed to generate real-world mixed-type financial tabular data for a variety of downstream tasks, for example, economic scenario modeling, stress tests, and fraud detection. The model uses embedding encodings to model mixed modality financial data, comprising both categorical and numeric attributes. The performance of FinDiff in generating synthetic tabular financial data is evaluated against state-of-the-art baseline models using three real-world financial datasets (including two publicly available datasets and one proprietary dataset). Empirical results demonstrate that FinDiff excels in generating synthetic tabular financial data with high fidelity, privacy, and utility.
- 25 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 64-72
Open AccessFinDiff: Diffusion Models for Financial Tabular Data Generation
The sharing of microdata, such as fund holdings and derivative instruments, by regulatory institutions presents a unique challenge due to strict data confidentiality and privacy regulations. These challenges often hinder the ability of both academics and practitioners to conduct collaborative research effectively. The emergence of generative models, particularly diffusion models, capable of synthesizing data mimicking the underlying distributions of real-world data presents a compelling solution. This work introduces Financial Tabular Diffusion (FinDiff), a diffusion model designed to generate real-world mixed-type financial tabular data for a variety of downstream tasks, for example, economic scenario modeling, stress tests, and fraud detection. The model uses embedding encodings to model mixed modality financial data, comprising both categorical and numeric attributes. The performance of FinDiff in generating synthetic tabular financial data is evaluated against state-of-the-art baseline models using three real-world financial datasets (including two publicly available datasets and one proprietary dataset). Empirical results demonstrate that FinDiff excels in generating synthetic tabular financial data with high fidelity, privacy, and utility.
- Open Access
- 4 July 2023
- Published by Elsevier in SSRN Electronic Journal
Assuring Sustainable Futures: Auditing Sustainability Reports using AI Foundation Models
Ensuring compliance of sustainability reports, especially Environmental, Social, and Governance (ESG) reporting, with the EU Taxonomy is essential in the European Union. In parallel, a significant transformation of Artificial Intelligence (AI) can be observed, shifting from task-specific models towards Foundation Models. This study investigates the incorporation of AI Foundation Models into the auditing process of ESG reports. OpenAI’s GPT, a foundation model, is utilized in this study to analyze sustainability reports to assess their compliance with the EU Taxonomy and providing auditors with critical insights. The integration of Foundation Models with traditional auditing methodologies signifies the advent of an innovative epoch in auditing of ESG reports, characterized by enhanced efficiency and insights.
- 4 July 2023
- Published by Elsevier in SSRN Electronic Journal
Open AccessAssuring Sustainable Futures: Auditing Sustainability Reports using AI Foundation Models
Ensuring compliance of sustainability reports, especially Environmental, Social, and Governance (ESG) reporting, with the EU Taxonomy is essential in the European Union. In parallel, a significant transformation of Artificial Intelligence (AI) can be observed, shifting from task-specific models towards Foundation Models. This study investigates the incorporation of AI Foundation Models into the auditing process of ESG reports. OpenAI’s GPT, a foundation model, is utilized in this study to analyze sustainability reports to assess their compliance with the EU Taxonomy and providing auditors with critical insights. The integration of Foundation Models with traditional auditing methodologies signifies the advent of an innovative epoch in auditing of ESG reports, characterized by enhanced efficiency and insights.
- 1 January 1998
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2419-2424 vol.4
ERSO-acquisition, reconstruction and simulation of real objects
A basic system for acquisition, reconstruction and simulation of real objects (ERSO) is presented. Our approach is to combine the knowledge of different research areas as photogrammetric, computer graphics, and computer vision to develop new techinques for generating three-dimensional models from images. The first two paragraphs give an introduction and overview of the system architecture. The following paragraphs describe in more detail the different parts of the reconstruction process
- 1 January 1998
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2419-2424 vol.4
ERSO-acquisition, reconstruction and simulation of real objects
A basic system for acquisition, reconstruction and simulation of real objects (ERSO) is presented. Our approach is to combine the knowledge of different research areas as photogrammetric, computer graphics, and computer vision to develop new techinques for generating three-dimensional models from images. The first two paragraphs give an introduction and overview of the system architecture. The following paragraphs describe in more detail the different parts of the reconstruction process