Generative AI Agents for Knowledge Work Augmentation in Finance

Recent advances in generative AI and machine learning have enabled systems like ChatGPT to produce increasingly human-like text.ChatGPT and its successors represent an exciting opportunity to expand human creativity and productivity, but they also pose real risks and concerns that must be addressed.These generative AI systems are able to generate plausible paragraphs, articles, images, and more on any topic or in any style based on analyzing massive amounts of data.They can complement and augment human abilities, helping people write faster, develop new creative ideas, and produce content at scale.By enabling AI agents to take over more routine and repetitive forms of manual work, human lives may be enriched with increased leisure time and opportunity for meaningful work.However, there are valid worries about bias and unfairness being reinforced or even amplified in generative AI models due to flaws in training data or algorithms.There are also concerns that these systems could be used to generate spam, scams, fake content, and manipulate public opinion at scale.As the technology improves, there may be concerns about AI systems developing their own objectives and goals that differ from human values and priorities.While ChatGPT and similar models represent an exciting new frontier of capability, they must be developed and applied carefully and ethically.Partnerships between humans and AI agents could help maximize the benefits of generative AI while mitigating the risks, but it will require proactive safeguards and oversight.Policies and guidelines should aim to ensure that generative AI progresses in a way that is aligned with human judgment, enhances trust in information, and supports human dignity and well-being.With proper stewardship, generative AI can positively transform our world, but we must be vigilant and thoughtful in how we develop and deploy this powerful technology.The opportunity of ChatGPT comes with real responsibilities that cannot be ignored.With concerted action, the future of generative AI can be shaped for the benefit of humanity.

The rapid advancements in generative AI have led to text-to-video models creating highly realistic content, raising serious concerns about misinformation spread through synthetic videos. As these AI videos become more convincing, they threaten information integrity across social media, news, and digital communications. Using AI-generated videos, bad actors can now create false narratives, manipulate public opinion, and influence critical processes like elections. This technology's democratization means that sophisticated disinformation campaigns are no longer limited to well-resourced actors, creating an urgent need for reliable detection methods and human-machine cooperation to maintain public trust in visual information across our digital transformation landscape. The accessibility of these tools to a broader audience amplifies the potential for widespread misinformation, making robust detection systems crucial for maintaining social media integrity.Through our research into video generation models, we identified that state-of-the-art systems like diffusion transformers operate on patches of noisy latent spaces. We deliberately mirrored this architecture in our classifier design, enabling it to analyze videos using the same fundamental structural unit generation models used to create them. This architectural alignment allows our system to adapt to emerging generation techniques while maintaining detection efficacy.We designed an explainable video classifier using deep learning and neural networks that detect AI-generated content and show evidence for its decisions. The classifier uses three main parts: a convolutional encoder that turns video frames into latent representations, a patch vectorizer that breaks these representations into analyzable chunks, and a transformer that processes these chunks to make the final decision. This human-centered computing design lets us efficiently process videos while maintaining explainability through Integrated Gradients, which reveal which input parts influenced the model's decisions.We use integrated gradients to show which parts of a video led to the model's decision. This method looks at how the model's decision changes as we move from a blank video to the actual video, showing us which pixels matter most for classification. These pixel-level maps provide clear evidence of why the model thinks a video is AI-generated or real, providing transparency critical for building trust in automated content verification systems.We will test our model on the GenVideo dataset, a comprehensive collection of videos labeled as real or AI-generated from diverse sources, including Stable Diffusion, Sora, Kinetic 400, and MSRVTT. This large-scale data analytics evaluation will check how well it classifies videos and explains its decisions, helping determine if the model can work as a practical tool for machine learning-based content verification, considering that wrong AI classifications could harm content creators' reputations.Our work adds to the growing field of explainable AI in content authentication and shows why we need clear evidence when making high-stakes decisions about video content. Future work will look at detecting hybrid videos (real videos with AI elements added) and making our visual explanations more useful for human decision-makers in content verification. The insights gained will inform the development of more sophisticated detection systems capable of addressing evolving challenges in digital content authenticity.

Machine learning has surged in popularity in recent years, with deep neural networks (DNNs) becoming a cornerstone of applications such as autonomous driving, augmented reality, and natural language processing, owing to their remarkable accuracy and scalability. As the demand for greater portability and flexibility grows, there is a strong push to optimize these applications for resource-constrained edge devices like smartphones and IoT devices.However, deploying DNN models on such devices presents significant challenges, primarily due to the large size of these models, which is a key contributor to their high performance. For instance, the DeepSpeech 2 model outperforms its predecessor, DeepSpeech 1, but its size is eight times larger, encompassing approximately 68 million parameters. Similarly, other networks designed for various tasks, including convolutional neural networks like VGG and transformer-based networks like BERT and GPT, also exhibit substantial sizes. For example, the GPT-3 model, a predecessor to ChatGPT, consists of an astonishing 175 billion parameters. Consequently, fitting DNN models onto resource-limited edge devices is a formidable challenge. My thesis addresses these challenges with a three-part approach: * High-Performance DNN Accelerators: I will develop DNN accelerators that excel in high performance, low latency, and energy efficiency through an algorithm and hardware co-design across diverse FPGA architectures (Chapter 2). * Real-Time DNN Execution: I aim to achieve exceptional accuracy and real-time execution for various large-scale DNNs on off-the-shelf mobile devices, IoT devices, and other platforms (Chapter 3). * CompVQC Framework: I will explore a systematic framework, CompVQC, to reduce the circuit length of quantum neural networks (QNNs) through model compression, which can achieve up to a 20% improvement in accuracy. This reveals that compilation-aware compression can enhance the robustness of QNNs on near-term noisy quantum devices (Chapter 4). The field of AI has recently undergone a significant transformation, ushering in the era of generative AI, where models like ChatGPT and Stable Diffusion create realistic and creative content for a wide range of applications. As generative AI becomes more prevalent, it is crucial to adapt algorithm-hardware co-design strategies to meet the specific needs of these models, particularly on edge devices. In the final chapter (Chapter 5), I will introduce my ongoing and future work focused on activating generative AI on the edge.--Author's abstract

New paradigms of machine processing are made possible by recent advancements in AI, which allowed for a transition from data-driven, discriminative AI jobs to complex, creative tasks through generative AI. Generative AI is an exciting & quickly developing field with enormous potential for problem-solving, data synthesis, & creative expression. GANs have emerged as an AI research hotspot. GANs are based on theory of adversarial learning and consist of a discriminator & generator. Estimating possible distribution of actual data samples & creating new samples are objectives of GANs. GANs are extensively researched because of their numerous potential applications (voice & language processing, image & vision computing, etc). The chapter provides clarification on the notion of generative AI & emphasizes its significance in the context of AI. It offers an in-depth exploration of different methods employed in generative AI. It analyzes current state of the art in GANs & provides an outlook for future. It draws attention to the field's present shortcomings & possible future advancements.

Streamlining Radiology Reporting: A Hands-Free Approach With Voice to Text and Generative Artificial Intelligence

In this special issue, we explore the opportunities and challenges of using Generative AI (GenAI), in particular, text generators in higher education learning and teaching. As GenAI has become increasingly popular with many staff and students, this special issue provides an overview of the current state of the field and offers insights into future research. This introduction paper consists of four parts. It begins by providing an overview of AI and Generative AI, identifying the gap and framing the special issue relating to the gaps. The second part explores the opportunities and challenges of GenAI in higher education, as identified in the literature. The third part provides an overview of the papers included in the special issue. The final part is the self-reflection of the lead author. The special issue aims to serve as a valuable resource for higher education stakeholders, such as students, practitioners, researchers and managers. We hope this collection will help advance knowledge and future research, encourage innovation and inform evidence-based policy and practices in the field of Generative AI in higher education.

<p><strong>Because of its wide range of applications, generative artificial intelligence (Generative AI) has received a lot of attention in academia and industry. Despite significant advances in computer vision and natural language processing, the use of Generative AI in tabular data is still underexplored. This gap is especially significant given the prevalence of tabular data as the primary data modality. This thesis seeks to close this gap by focusing on the efficient generation, evaluation, refinement, and application of tabular data, while addressing the challenges inherent in its heterogeneous nature.</strong></p><p>The primary goal of this thesis is to develop efficient algorithms for tabular data synthesis, advancing the field of Generative AI in the tabular domain. To achieve this goal, four specific research objectives have been outlined.</p><p>First, this thesis addresses gaps in evaluation metrics by unifying a framework for the comprehensive and consistent assessment of synthetic data. The proposed framework is designed to meet diverse business requirements across various downstream tasks by incorporating a wide range of advanced metrics. These metrics cover various types of evaluations, including univariate, bivariate, multivariate, cluster, and record-level evaluations. Additionally, standardized visualizations are provided to facilitate qualitative assessments. The results indicate that the proposed evaluation framework not only enables consistent comparisons, rankings, and the selection of different data synthesis approaches but also acts as a valuable tool to promptly assess the reliability of results obtained from synthetic data.</p><p>Second, this thesis presents innovative algorithms for the generation of tabular data, aiming to significantly enhance the quality of data synthesis. One of the contributions is the development of a reversible feature engineering pipeline designed to automatically represent tabular data in an efficient format while also ensuring that the transformed data can be easily converted back to its original format. Additionally, the thesis proposes novel deep learning-based tabular data generation models that are capable of learning the joint distribution of multivariate datasets without relying on predefined distribution assumptions. Experimental results highlight the effectiveness of the proposed algorithms in handling heterogeneous data, demonstrating superior performance in most scenarios within the same training duration when compared to similar alternatives.</p><p>Third, this thesis introduces innovative synthetic data prototype selection algorithms with the goal of refining the generated samples. This approach aims to leverage the advantages of Deep Generative Models, which, once trained, can generate unlimited and diverse synthetic data. By carefully identifying and selecting high-quality samples or removing unrealistic ones, the quality of the synthetic data can be enhanced from a post-processing perspective. Building on this hypothesis, we recognize the iterative nature inherent in the data synthesis procedure, where the processes of data generation, evaluation, and refinement should operate repeatedly in a cyclical flow. During the data generation phase, the evaluation framework assists in identifying potential issues and risks based on use cases, while the refinement (post-processing) step iteratively improves synthetic data in alignment with the evaluation outcomes.</p><p>Last, this thesis applies and validates the proposed models across various domains, including business, healthcare, and government, each requiring distinct downstream tasks. Specifically, we assess our algorithms from the standpoint of data balancing using churn data, evaluate our models focusing on data augmentation with health data, and test our algorithms from the perspective of data representation considering data privacy concerns within sensitive citizen data. These applications serve as practical illustrations, demonstrating the effectiveness and utility of the proposed Generative AI models in real-world scenarios.</p><p>In summary, this thesis makes a substantial contribution to the advancement of Generative AI within the domain of tabular data. It not only presents innovative algorithms and evaluation methods but also introduces practical frameworks applicable to real-world scenarios. The findings demonstrate the considerable capacity of Generative AI to fundamentally transform tabular data in various fields, ultimately leading to improved data availability, quality, and quantity.</p>

BackgroundGenerative artificial intelligence (GenAI) is rapidly entering knowledge work, yet organizational psychology lacks a clear account of when and how GenAI functions as a job resource in the Job Demands–Resources (JD-R) model, especially in high-demand academic work.MethodsWe conducted a JD-R–guided directed qualitative content analysis of a de-identified human–AI interaction log generated during routine academic work. The log was segmented into interactional episodes and coded using function-first descriptors. Codes were mapped deductively to three JD-R–aligned resource domains (cognitive; structural/strategic; emotional/psychosocial) and a boundary-conditions stream (human oversight; data integrity/traceability).ResultsAcross 15 episodes, GenAI enacted cognitive functions that reduced informational complexity, structural/strategic functions that increased planning capacity and task structure, and emotional/psychosocial functions expressed through observable efficacy-reinforcing and action-orienting cues. Boundary-condition coding showed that benefits depended on human oversight and integrity routines; in episodes requiring substantial verification or traceability work, GenAI could shift rather than reduce demands.ConclusionGenAI can operate as a conditional job resource in demanding academic knowledge work, but sustainable benefit requires explicit human-in-the-loop oversight and data integrity practices that support reliable and responsible use.

Understanding how AI influences employee emotions is becoming critical as organizations prepare for widespread AI agent deployment. While existing research has explored human-AI interactions in corporate settings, little is known about how employees emotionally navigate relationships with AI agents exhibiting distinct personality traits. This empirical study examines white-collar employees' emotional responses while interacting with three generative AI agents in a virtual workspace, revealing novel social dynamics enabled by AI technologies. Using qualitative methods and inductive analysis, our findings show that anthropomorphic AI agents evoke a broad spectrum of emotions, from connection and contentment to amusement and frustration , extending beyond those typically triggered by web-based AI agents. Notably, participants experienced new emotional subsets, including unique manifestations of relational assurance and perceived worthlessness , which introduces new emotional subcategories within established frameworks. Moreover, the visual embodiment of AI agents in virtual workspaces significantly shapes user expectations and satisfaction. While a more human-like appearance can enhance engagement, it also introduces risks—a mismatch between an AI’s visual representation and its actual behavior can heighten disappointment if the AI fails to meet human-like expectations. As organizations integrate AI agents into the workplace, our findings provide key insights for designing effective human-AI interactions. We emphasize the importance of human-centered design approaches that foster, rather than hinder, employee engagement, ensuring AI contributes positively to corporate environments. • Presents an empirical study exploring white-collar employees' interactions with anthropomorphic AI agents in a virtual workspace, offering actionable insights for human-centered corporate AI design and deployment. • Extends a framework for understanding emotional dynamics in human-AI interactions, with a particular focus on emerging emotional patterns unique to anthropomorphic AI collaboration. • Explores the innovative convergence of AI and VR technologies, laying the groundwork for research on future workspaces where human employees collaborate with virtual AI agents.

The emergence of autonomous AI agents—systems proficient in reasoning, planning, and executing intricate tasks with digital tools—signifies a pivotal transformation in automation. In contrast to conventional generative AI, these agents function with considerable autonomy, offering potential productivity enhancements in fields such as healthcare, finance, and education. Nevertheless, their autonomy presents new governance difficulties, encompassing liability, monitoring, and systemic hazards. This study delineates the capabilities of AI agents, assesses their societal ramifications, and proposes a dynamic, evidence-informed governance system. Policymakers should leverage the advantages of AI agents and mitigate dangers through regulatory sandboxes, transparency standards, and international coordination. Our plan prioritizes collaboration among governments, industry, and civil society to guarantee safe, equitable, and innovative deployment of agents.

The TrollLabs Open dataset includes comprehensive information that offers a comparison of design practices and outcomes between human participants and Generative AI during a hackathon event. The dataset was curated through the running of a prototyping hackathon designed to assess the abilities and performance of generative AI, specifically ChatGPT, in the early stages of engineering design. This assessment involved comparing ChatGPT's performance to that of experienced engineering students in a hackathon setting, where participants competed by making a prototype that fires a NERF dart as far as possible. In this setup, all ideas, concepts, strategies, and actions undertaken by the AI-controlled team were autonomously generated by the ChatGPT, without human intervention or guidance, but implemented by two participants. Five self-directed baseline teams competed against the AI team. The dataset comprises 116 prototype entries and 433 edges (connection) that enable comparative analysis of design practices and performance between the team instructed solely by generative AI and baseline teams of experienced engineering design students. Prototypes and their attribute data were captured using Pro2booth, an online prototype capture platform running on participants' phones and computers. The dataset includes a transcript of the conversation between ChatGPT and the team responsible for implementing its recommendations, featuring 97 exchanges of prompts and responses. It contains the initial prompt used to instruct the AI, the objective and rules of the hackathon and the objective performance of teams, showing the ChatGPT team finishing 2nd among six teams. To the authors' knowledge, the TrollLabs Open dataset is the first and only open resource that directly compares the performance of generative AI with human teams in an engineering design context. Thus, it is intended to be a valuable resource to design researchers, engineering and design students, educators, and industry professionals seeking to find strategies for implementing generative AI tools in their design processes. By offering a comprehensive data collection, the dataset enables external researchers to conduct in-depth analyses that could highlight the practical implications of integrating generative AI in design practices, possibly providing an overview of its limitations and presenting recommendations for improved integration in the design process.

Purpose Recent advances in generative AI (GenAI) are transforming multiple aspects of society, including education and foreign language learning. In the context of English as a Foreign Language (EFL), significant research has been conducted to investigate the applicability of GenAI as a learning aid and the potential negative impacts of these technologies. However, there are few formalised frameworks available to support the integration and development of AI literacy skills for EFL learners. Design/methodology/approach In this conceptual article, we demonstrate the way in which an existing framework, which is designed for integration of AI into educational assessment (the AI Assessment Scale), can be adapted to structure classroom activities, curricula and assessment in the context of EFL writing and translation. To achieve this, we contextualise the approach by firstly reviewing the extant literature on GenAI and EFL writing. Following this, we focus on Levels 2, 3 and 4 of the AI Assessment Scale (AIAS) and explore possible pedagogical adaptations to suit an EFL writing context. At the same time, we reflect on the possible advantages and limitations of this approach. Findings We argue that for pedagogy to succeed in a digital world, educators in the EFL field must not turn away from the affordances offered by AI technologies. At the same time, abundant caution is needed given the lack of empirical data on how these technologies may impact learning in the long term. To this end, we contend that a balanced approach, which clearly describes and communicates how AI may be used in learning activities, can be beneficial for both educators and learners in the EFL classroom. Practical classroom implications are included for writing and translation so teachers can apply the patterns immediately. Originality/value The adaptation of the AIAS from an assessment redesign tool to a framework for structuring EFL pedagogy is the core contribution and innovation in this article. The use of this framework has significant implications for classroom practice in the age of AI, as it enables a structured redesign of existing pedagogical activities in light of new technologies.

Abstract: Augmented Human Intelligence (AHI) represents a paradigm shift in human-AI collaboration, leveraging Generative AI, Quantum Computing, and Extended Reality (XR) to enhance cognitive capabilities, decision-making, and immersive interactions. Generative AI enables real-time knowledge augmentation, automated creativity, and adaptive learning, while Quantum Computing accelerates AI optimization, pattern recognition, and complex problem-solving. XR technologies provide intuitive, immersive environments for AI-driven collaboration, bridging the gap between digital and physical experiences. The convergence of these technologies fosters hybrid intelligence, where AI amplifies human potential rather than replacing it. This research explores AI-augmented cognition, quantum-enhanced simulations, and AI-driven spatial computing, addressing ethical, security, and societal implications of human-machine synergy. By integrating decentralized AI governance, privacy-preserving AI techniques, and brain-computer interfaces, this study outlines a scalable framework for next-generation augmented intelligence applications in healthcare, enterprise intelligence, scientific discovery, and immersive learning. The future of AHI lies in hybrid intelligence systems that co-evolve with human cognition, ensuring responsible and transparent AI augmentation to unlock new frontiers in human potential. Keywords: Augmented Human Intelligence, Generative AI, Quantum Computing, Extended Reality, XR, AI-driven Cognition, Hybrid Intelligence, Brain-Computer Interfaces, AI Ethics, AI-enhanced Learning, Spatial Computing, Quantum AI, Immersive AI, Human-AI Collaboration, Ethical AI Frameworks.

Visual learning analytics (VLA) is becoming increasingly adopted in educational technologies and learning analytics dashboards to convey critical insights to students and educators. Yet many students experienced difficulties in comprehending complex VLA due to their limited data visualisation literacy. While conventional scaffolding approaches like data storytelling have shown effectiveness in enhancing students’ comprehension of VLA, these approaches remain difficult to scale and adapt to individual learning needs. Generative AI (GenAI) technologies, especially conversational agents, offer potential solutions by providing personalised and dynamic support to enhance students’ comprehension of VLA. This controlled lab study investigates the effectiveness of GenAI agents, particularly when integrated with scaffolding techniques, in improving students’ comprehension of VLA. A randomised controlled trial was conducted with 117 higher education students to compare the effects of two types of GenAI agents: passive agents , which respond to student queries, and proactive agents , which utilise scaffolding questions, against standalone scaffolding in a VLA comprehension task. The results show that passive agents yield comparable improvements to standalone scaffolding both during and after the intervention. Notably, proactive GenAI agents significantly enhance students’ VLA comprehension compared to both passive agents and standalone scaffolding, with these benefits persisting beyond the intervention. These findings suggest that integrating GenAI agents with scaffolding can have lasting positive effects on students’ comprehension skills and support genuine learning. • Low visualisation literacy limits understanding of visual learning analytics (VLA). • Investigated the effects of GenAI agents with scaffolding on VLA comprehension. • Conducted RCT comparing proactive, passive AI agents, and standalone scaffolding. • Proactive GenAI agents significantly improved VLA comprehension with lasting effects. • GenAI agents offer sustainable improvements in students’ comprehension skills.

The use of Conversational AI (CAI) agents within education has seen a rise with the rapid integration of generative AI (GenAI). The generative ability of the application combined with conversational capabilities has enhanced the perceived and actual usefulness of CAI applications. Given this development, it is critical to undertake a comprehensive review to understand the actual application domains, challenges, and efforts within this area. A range of empirical studies as well as reviews have been undertaken in recent years, but the current understanding remains fragmented. To better understand the current state-of-the-art, current trends and future implications of CAI on education, we conducted an umbrella review (UR) to systematically synthesize findings from thirty-four review articles. Articles were collected through a search across five major databases. They were screened using predefined eligibility criteria focusing on CAI agents used across educational domains and contexts. The PRISMA framework for transparent reporting is followed throughout the process and a thematic analysis has been undertaken to analyze the data. The results show that CAI utilization is concentrated in pedagogical applications such as teaching support, psychological engagement, and metacognitive development, while administrative functions, research assistance, and specialized training remain underdeveloped. Technical limitations and concerns with educational impact dominate discussions. Ethically, human-AI relationship concerns persist across all CAI generations, while academic integrity and data privacy represent emerging areas of concern. The review reveals gaps in CAI frameworks: lack of end-to-end design guidance, weak CAI specific usability methods, unclear pedagogical guidance and classroom implementation strategies, and limited AI literacy support. The article concludes by proposing a roadmap for ethical CAI implementation in education and identifying priority areas for future research.