When generative artificial intelligence (AI) was released to the public in late 2022, it paved the way for new and disruptive solutions to enter the corporate world. Boasting extraordinary potential for a range of business functions, financial institutions have already begun integrating the technology into their existing systems and processes.
In this article, we look to cut through the hype and negative noise surrounding AI by demonstrating how Large Language Models (LLMs) like ChatGPT can be integrated with process automation platforms to provide compelling, enterprise grade solutions to some of the financial sector’s biggest challenges. Not least, in the arena of regulatory intelligence, where it is allowing users to capture and interact with multiple regulatory news bulletins, produce summaries for distributions to end clients, and more.
Building on the middle ground through integration
Many organizations have developed their own automation initiatives within AI Centers of Excellence. In addition, the arrival of specialist model providers and generalist models provided by hyperscalers, like AWS (Textract) and Azure (Form Recognizer), has created a middle ground rife with both uncertainty and opportunity.
For finance and operations, this middle ground has been particularly fraught. As a highly regulated area with stringent demands for repeatability, consistency, transparency, and high levels of control, it is difficult to bring AI projects into production. Consequently, enterprise platforms that offer contained and integrated AI environments can help AI Centers of Excellence get to value more quickly. The platforms accomplish this by providing safe and quick access to AI models and coupling them with enterprise automation features that enhance the likes of transformation and workflows.
Although the ways in which modern AI and ML can be applied to business processes is theoretically limitless, right now, one of the most immediate relates to regulatory intelligence.
Applying AI to regulatory intelligence
In the data automation space, LLMs can capably deal with scenarios where the layout and format of data sources change significantly and often. However, their use in data pipelines – where criticality of accuracy and repeatability is imperative – is less reliable. For now, this precludes publicly available AI models from being used freely across all business functions.
However, regulatory intelligence is different. On a daily basis, leading tax consultancies, regulators, and local market experts release news bulletins that cover changes to global tax regulations. Received as emails or PDFs, tax professionals are presented with an abundance of information that must be manually interrogated to identify which changes will affect their core tax processes. These changes must then be summarized, distributed downstream to clients, and then implemented into client systems. The impact on time and resources is keenly felt across the industry. It is a friction point within tax processing that has, until now, been denied any kind of practical solution. By integrating platforms with AI models, the promise of such a solution is finally within reach.
AI can be used to capture new tax regulation at source and examine the contents on behalf of the tax professional. Rather than spending hours poring over documents, tax professionals can simply direct the AI to prioritize the outreach in order of importance, extract the relevant information, and automatically produce summaries.
Automated alerting and reporting that captures regulatory news bulletins is delivered through email notifications, dashboards, or other forms of communication and highlights important updates, trends, and potential impacts on the business. By using event extraction techniques, AI can then detect and track specific regulatory events or changes mentioned in the news bulletins, such as policy updates, regulatory announcements, enforcement actions, or changes in compliance requirements.
The tracking of these events means keeping up-to-date with the latest regulatory developments is quicker and easier. Moreover, regular performance evaluations of your AI system will allow you to make improvements based on user feedback, while new data helps to further refine the system’s accuracy and ensure it remains current with an evolving regulatory landscape.
AI as a tool, not a replacement
No one can say with any degree of certainty how AI will come to transform our understanding of the nature of work in the coming years. However, right now, tools such as ChatGPT are simply not capable of performing vital tasks unless they’re embedded in a controlled end-to-end business process that supports scalability, for example, by monitoring more news sources or reacting to updates faster.
As such, where AI is integrated with platforms for processes such as regulatory intelligence, it is not to replace human workers, it is to provide them with a formidable tool.
Hours spent each day scrutinizing outreach on new regulations can be reduced to minutes, leaving this newly freed time to be exploited for more value-adding activities. Governance is built-in too. In collaboration with IT teams, system prompts developed by tax professionals can prime AI to respond in certain ways, thereby preventing the risk of users venturing beyond the tool’s designated boundaries.
This level of control equips AI with a human-like attitude to work, meaning it will answer questions within supplied contexts and request clarification rather than provide responses based on guesswork. As businesses and organizations continue to adopt AI technologies, it is crucial that they develop working practices to ensure the limits of AI are not exceeded. The system prompt provides a clear mechanism to set these boundaries, eliminating the risk of misuse and ensuring accountability.
In time, platforms will inevitably go on to integrate with multiple AI systems, all capable of completing different tasks within the whole regulatory space. Organizations will be able to decide which use cases different systems are best suited to and manage how users operate each one. Judgements that were once the preserve of IT teams will be democratized across all operational departments within the organization.
In the meantime, when integrated into platforms, ChatGPT and other adjacent LLMs are set to provide transformational empowerment to those operating in the regulatory intelligence space.
Nico Busch, Product Manager, Xceptor
Nico Busch joined Xceptor just over a year ago as Senior Product Manager for the Xceptor Platform. He graduated with a master’s degree in risk management and financial engineering at Imperial College London, and a bachelor’s degree in philosophy and economics at the London School of Economics and Political Science. With 18 years’ experience in product development and product management in the finance and technology industries, Nico has the knowledge and ability to help our clients to confront their business challenges and optimize their company’s performance. Trusted by the world’s largest banks, the Xceptor platform is highly configurable and elevates operational efficiency, enhancing critical business processes.