The CRAFiC workshop 'AI in finance and accounting: Opportunities, Challenges and Societal Implications' was funded by the Research and Knowledge Exchange/Impact Stimulation Fund and brought together academics working at the frontier of finance, accounting, artificial intelligence, corporate disclosure, ESG, market behaviour and governance. The aim was not only to present current research papers, but also to begin shaping a broader research agenda around responsible AI, market resilience, disclosure credibility, and the future of finance and accounting.
The motivation for the workshop was simple but important. AI is no longer a distant possibility for finance and accounting. It is already being adopted in investment, corporate decision-making, financial communication, risk assessment, asset management and organisational processes. The central question is therefore not whether AI will be adopted, but how it will be used, who benefits from it, what risks it creates, and how firms, markets, regulators and researchers should respond.
The day was structured around two main themes. The morning session, chaired by Dr Jiao Ji, focused on AI, Financial Stability, and Market Transformation. The afternoon session, chaired by Dr Matt Burke, turned to AI, Governance, Accountability, and Organisational Change. The formal programme was complemented by a PhD poster exhibition, informal networking, and a closing roundtable discussion on responsible AI in finance and accounting, moderated by Dr Konstantinos Tolikas.
The workshop opened with Professor Fangming Xu’s keynote presentation on machine learning and M&A strategies. Professor Xu, from the University of Bristol, examined whether machine learning can improve the prediction of takeover premiums using information available before merger announcements. By combining accounting fundamentals, deal and advisor characteristics, and macro-financial indicators, the paper showed how machine learning methods can provide a more systematic and predictive framework for screening and benchmarking M&A premiums. Importantly, the presentation also highlighted the role of explainable AI, reminding us that prediction alone is not enough: in finance, models also need to be interpretable, economically meaningful and useful to decision-makers.
The morning then moved to ESG disclosure and greenwashing. Dr Helen Lu, from Vlerick Business School, presented “Specifics matter: An analysis of mutual fund ESG disclosures.” Her paper examined the specificity of ESG disclosures in mutual fund prospectuses and asked whether more concrete disclosures contain meaningful information about future fund behaviour. The findings showed that specificity matters: more specific ESG disclosures are associated with future ESG-related portfolio choices and investor responses. The paper raised important questions about the quality of ESG communication and the distinction between generic language and decision-useful disclosure.
Dr Shuo Wang, from the University of Edinburgh Business School, then presented work on corporate greenwashing accusations as signals of climate credibility gaps. This paper used large language models to classify different types of media-reported greenwashing accusations and linked these accusations to firm-level climate performance. A key message was that media scrutiny can provide timely signals of possible gaps between corporate climate talk and corporate climate action, but these signals are selective. The paper showed how AI can be used as a research tool to organise and interpret large volumes of textual information, while also contributing to the broader debate on climate accountability.
The final morning paper, presented by Professor Álvaro Cartea, from the University of Oxford, examined AI bubbles with large language models. This paper asked whether AI agents participate in speculative bubbles and whether more sophisticated AI necessarily improves market efficiency. The findings were striking. AI agents can engage in speculative trading even in environments where theory predicts no trade, and greater reasoning capacity does not necessarily eliminate bubbles. The paper therefore challenged a common assumption that more advanced AI will automatically make markets more efficient. Instead, it suggested that AI agents may reproduce, and in some settings amplify, behavioural and strategic frictions that are familiar from human markets.
The afternoon began with Professor Paul Geertsema’s keynote address on “Agentic AI: promise, pitfalls and practical reality”. Professor Geertsema, from Vlerick Business School, broadened the discussion from individual research papers to the wider implications of agentic AI for organisations, decision-making and professional work. The presentation highlighted both the excitement surrounding AI agents and the need for caution about implementation, accountability, robustness and organisational readiness. A recurring theme was the gap between technological promise and practical reality: AI systems may be powerful, but their value depends on how they are deployed, monitored and governed.
Professor Spyros Galanis, from Durham University, then presented “Information Aggregation with AI Agents.” The paper studied whether AI agents can aggregate dispersed private information through prediction markets. The findings suggested that AI agents can perform well in relatively simple information environments, but their ability to aggregate information deteriorates as complexity increases. This has important implications for the use of AI in markets and multi-agent environments. If AI agents struggle with higher-order reasoning, private information, or strategic inference, then the design of markets and institutions involving AI agents becomes a central research and policy issue.
Dr Chelsea Yaqiong Yao, from Lancaster University, presented “Beyond Greenwashing: ESG Disclosure Substance, Investor Flows, and Fund Behavior.” The paper examined ESG disclosure in mutual fund shareholder letters and showed that the content of ESG disclosures is central to understanding both investor responses and subsequent fund behaviour. Rather than treating ESG disclosure as a single category, the paper distinguished between different forms of ESG communication and showed that investors and fund managers respond differently depending on the substance of the disclosure. This reinforced one of the workshop’s broader messages: language matters, but the content, credibility and consequences of language matter even more.
The final paper presentation was by Silun Wang, a PhD candidate at Sheffield University Management School, on AI talk, AI investment and governance. Her work examines “AI washing” as the gap between firms’ AI narratives and observable AI-related commitments. The paper studies how governance features, including managerial ownership, institutional ownership, board structure and CEO duality, shape firms’ propensity to engage in AI washing. This presentation connected directly with the workshop’s governance theme. As AI becomes a dominant corporate narrative, investors and regulators face the challenge of distinguishing credible AI adoption from exaggerated or opportunistic claims.
Across the day, several cross-cutting lessons emerged. First, AI is both a tool for research and an object of research. Some papers used AI and machine learning to analyse complex financial and textual data. Others studied how AI agents behave in markets, how firms talk about AI, or how AI may affect governance and accountability. This dual role makes AI especially important for finance and accounting scholars.
Second, the workshop challenged the idea that AI will automatically improve financial markets. The papers on speculative bubbles and information aggregation showed that AI agents may face limitations when reasoning strategically, interpreting the actions of others, or coordinating in complex environments. AI may improve prediction and information processing, but it may also generate new forms of opacity, mispricing, fragility and accountability problems.
Third, disclosure credibility emerged as a central issue. Several papers focused on the gap between what organisations say and what they do, whether in relation to ESG, climate commitments or AI adoption. As firms increasingly communicate through sustainability narratives, AI narratives and strategic disclosures, researchers need better tools to evaluate substance, specificity and verification. Boilerplate disclosure, selective communication, and “washing” behaviour are likely to remain major concerns for investors, regulators and society.
Fourth, responsible AI in finance and accounting requires interdisciplinary thinking. The workshop brought together ideas from corporate finance, asset management, accounting, governance, textual analysis, market microstructure, experimental economics and regulation. This reflected one of CRAFiC’s key strengths: the ability to connect financial markets research with broader questions of governance, accountability and societal impact.
The closing roundtable brought these themes together under the title “Responsible AI in Finance and Accounting: Research Priorities, Policy Challenges, and Future Directions.” The discussion focused on unresolved questions, research gaps, governance and regulation, financial stability, professional change and future collaboration. Participants reflected on whether AI applications are advancing faster than the evidence on their economic and societal consequences, how researchers can identify the actual use of AI inside firms and financial institutions, and whether existing regulatory and disclosure frameworks are sufficient for the risks created by AI.
The roundtable also looked forward. Possible next steps include developing collaborative research projects, exploring interdisciplinary funding opportunities, producing policy-oriented outputs, and building a CRAFiC research agenda around AI, responsible finance and market resilience. The discussion made clear that the workshop should not be seen as a one-off event, but as the beginning of a broader conversation.
We are grateful to all keynote speakers, presenters, discussants, PhD contributors and participants for making the day intellectually rich and constructive. Special thanks are also due to Tingting Tao and Guannan Hao for their excellent support with the workshop organisation, logistics, registration, catering, materials and speaker arrangements.
The central message from the day was clear. AI is already transforming finance and accounting, but its benefits will depend on how well researchers, firms, regulators and society understand and govern its risks. CRAFiC looks forward to continuing this conversation and developing future research on responsible AI, market resilience, disclosure credibility and the changing landscape of finance and accounting.