Artificial Intelligence in Financial Services 2025

To achieve these improvements and keep up with the rapid pace of innovation, business leaders will require a flexible AI strategy. They will also need to continuously monitor the technology landscape, especially over the next two to three years, as developments advance in key areas: –Small language models (SLMs): An SLM is a machine-learning algorithm that has been trained on a dataset that is smaller and more specific than those of large language models (LLMs). SLMs are typically deployed for a single specific task (e.g. answering customers’ questions about a particular product); they can be more efficient and faster than LLMs due to their size and their use of a more targeted dataset. –Retrieval-augmented generation (RAG): RAG is a process that enhances the capability of genAI models by optimizing LLMs – increasing both accuracy and reliability. This process uses in-house data repositories (e.g. internal organizational data such as policy or procedural documents) to support the validation of responses to ensure it is combing the latest and most accurate responses.6 Companies can incorporate RAG into their AI-based assistants, making them more effective and reliable at answering customer questions, by validating responses against a master repository (such as knowledge articles in a repository). –AI agents: These are designed and built to understand inputs, make decisions and act on them without any human intervention.7 When AI acts as an agent instead of a service, it is no longer locked into assisting with a single task or function, but can be applied to a broader range of problems and decision-making opportunities. An example of this is the use of AI agents that can process enquiries, action customer requests and make product recommendations.8 –Quantum computing: By integrating quantum computers with conventional supercomputers, parts of traditional problem-solving workflows can be accelerated. Leveraging quantum computing in AI models would allow faster processing of larger datasets and accelerated decision-making, especially in areas requiring complex pattern recognition. An example of this would be quicker identification of fraudulent transactions for banks and payment processors.9 As AI becomes central to technology strategies, executives must continually evaluate technology ecosystems to capture emerging opportunities, ensuring that AI investments are thoughtfully integrated into broader initiatives. Decisions by traditional financial institutions on partnerships or investments in financial technology companies – which are re-emerging as leaders in niche AI-driven solutions – will grow in importance.10 Striking the right balance between managing investment costs and risks, accelerating time to market and selectively building in-house solutions will be key to progressing towards scaled adoption of AI in financial services. Executives must continually evaluate technology ecosystems to capture emerging opportunities, ensuring that AI investments are thoughtfully integrated into broader initiatives. Artificial Intelligence in Financial Services 13