Author: Xavier Kasdan

Subject: Information System Science

Title: Benefits from implementing a data labelling tool

Abstract: 

The amount of data in the world is constantly increasing, making data management more complex and demanding. Effective data utilization is crucial for in-depth analysis, logical reasoning, and decision-making processes. Data labelling is an essential part of this process, but it has traditionally been labour-intensive and resource-consuming. To manage always scarce resources more efficiently, companies are turning to data labelling tools to automate the process, enhance data management, and extract more value from their data.

This thesis aims to reason the benefits and risks associated with implementing a data labelling tool, specifically Microsoft Purview. The study employs a benefit measurement model and includes a pilot project conducted in a case company. Additionally, interviews with company professionals were conducted to provide further validation and professional insights into the benefits of data labelling.

The findings reveal several notable benefits of data labelling and data labelling tools. Firstly, labelling tools improve the quality and understanding of the data in hand, enhancing its utility. Secondly, automated labelling tools significantly accelerate the labelling process, reducing resource consumption compared to manual methods. Thirdly, data labelling offers broad advantages in data management, data governance, data loss prevention, data security and compliance management and data lifecycle management. Risks related to data labelling tool implementation includes accuracy of labelling, user adoption and engagement and beneficial resource allocation.

Key words: Data labelling, Benefit management, Data labelling tool

Subject: International Master in Management of IT

Title: Enhancing Financial Audits through Deep Learning: Addressing Key Challenges and Improving Efficiency

Abstract: 

The financial audit (FA) process, traditionally based on manual procedures and reliant on professional judgment, faces challenges in the era of digitalization, due to the requirement of analyzing large volumes of complex data. This thesis investigates how deep learning (DL) can address challenges in the FA process, particularly focusing on large data volumes, manual procedures, and the subjectivity of professional judgment. Using the Task-Technology Fit (TTF) theory as a guiding framework, the study explores DL’s potential through a comprehensive research approach.

Through 13 EY expert interviews across various global locations, and a qualitative survey, the research identifies key challenges in current FA practices, and shows a fit with DL applications. DL shows promise in addressing these issues by automating tasks, managing data complexity and large data volumes, and providing auditors with data-driven recommendation.

Findings reveal that DL’s capabilities in natural language processing (NLP), computer vision, anomaly detection, recommendation systems, and big data analytics can address the identified FA challenges. Additionally, DL models are suggested for alleviating each challenge.

This study not only validates existing DL applications, but also introduces up to date FA challenges. This thesis provides a solid foundation for future research and practical applications in the field of financial auditing. The implications of these findings suggest that adopting DL can lead to more efficient and accurate FA processes.

Key words: Deep Learning (DL), Financial Audit (FA), Task-Technology Fit (TTF)

Subject: Information Systems Science

Title: Adoption and Governance of AI-Powered Dashboards in Executive-Level Decision-Making

Abstract: 

AI-powered dashboards are business intelligence tools that collect and present data in a comprehendible manner to enhance decision-making. The integration of artificial intelligence (AI) in data-driven decision-making (DDDM) processes is continuously increasing in modern organizations, however, current research examining AI-powered dashboards does not address the deployment of the technology. To complement prior research, this thesis aims to provide a foundation for the implementation of AI-powered dashboards by investigating the current dashboard best practices, as well as the adoption and governance of AI-powered dashboards in the context of executive decision-making. The concepts of DDDM and AI function as a foundation regarding the purpose and functionality of AI-powered dashboards, whereas IT governance provides a basis for their adoption and governance. The research methodology encompasses a review of scientific literature addressing these concepts to enable the proposal of two distinct theoretical frameworks addressing the adoption and governance of AI-powered dashboards.

The results of this study highlight three critical areas regarding AI-powered dashboards: current best practices, adoption processes, and governance. Best practices show that AI-enabled dashboards are dynamic and versatile BI tools that enhance decision-making and operational efficiency through detailed visualizations and data analysis. Concerning adoption, the study emphasizes the importance of selecting a suitable framework tailored to organizational needs, suggesting that a combination of existing models might often be necessary to integrate AI and BI within organizational environments effectively. Results addressing the governance of AI-powered dashboards emphasize the importance of BI and AI governance. Although the development of AI governance frameworks is still in its early stages compared to BI and IT governance, the findings suggest that adopting flexible and diverse governance structures enables organizations to manage the security, transparency, and accuracy risks associated with AI technologies. 

Key words: Dashboard, Business intelligence, Artificial intelligence, Data-driven decision-making.

Subject: Cybersecurity; Information Management

Title: Going passwordless – a case study on the implementation of passwordless at Accenture

Abstract: 

In recent years, passwordless authentication has become an increasingly popular topic in the cybersecurity industry. An increase in the amount of cyberattacks has warranted a need for better authentication methods, and passwordless is perfectly suited for this issue. As one of the first companies in the world, Accenture has implemented passwordless authentication at a large scale.

There is a lack of existing research into implementing authentication systems, especially into passwordless authentication. This thesis investigates the implementation of passwordless authentication at Accenture, through interviews with implementation team members, IT service desk members, and ‘regular’ employees, several key lessons for future implementations have been identified.

The study has concluded that clear and concise communication is a key factor in an implementation, as end-users are prone to misunderstanding or ignoring important communication. Furthermore, the IT service desk must be involved as a main stakeholder in an implementation, as they have a heavy burden to support the implementation. To address the gaps in existing research, future studies should focus on the IT service desk and conduct additional case studies to increase the knowledge on passwordless authentication.

Key words: passwordless, authentication, implementation, end-user communication, FIDO2, passkeys.