Barely a week can pass in financial services without a new AI innovation bursting onto the scene. While some show more promise than others, all share a mutual mission in trying to alleviate some of the industry’s biggest pain points.Agentic AI refers to creating intelligent AI “agents” that make autonomous decisions to carry out tasksA quick glance through some of the latest headlines will reveal applications in the areas of backend automation, cybersecurity, portfolio management, and personalisation.Notable recent developments include BNP Paribas’ multi-year partnership with Mistral AI, BBVA’s deployment of ChatGPT with OpenAI, and Franklin Templeton’s work in building an “advanced financial AI platform” with Microsoft.The industry’s appetite for AI is more ravenous than ever, and while the full possibilities presented by GenAI are still being discovered, agentic AI is emerging as a new focal point for the financial services industry.In a nutshell, Agentic AI describes creating intelligent AI “agents” capable of autonomous decision-making to carry out tasks, enabling businesses to completely automate historically arduous, lengthy, and labour-intensive processes.Here, FinTech Futures speaks with industry experts to lift the lid on agentic AI and its promises of seamless automation.Beyond data outputsIn the publication Artificial Intelligence: A Modern Approach, Peter Norvig and Stuart Russell cite an agent as “anything that can be viewed as perceiving its environment through sensors and acting upon that environment through actuators”.How does this differ from more traditional forms of AI? For Matt Roberts, head of data at ClearBank: “Agentic AI is not just about data outputs.”Roberts has led a team building AI and data infrastructure at ClearBank since 2022, and previously worked at Lloyds Banking Group for over a decade, predominantly in the field of data science.“It’s about taking actions and automating whole processes that might previously have been considered too complex due to the potential number of outcomes or the expert knowledge required to reach those outcomes. In contrast, traditional AI is very much focused on data outputs,” he tells FinTech Futures.Roberts says the concept of AI agents completing actions within an environment has been “supercharged” by two related developments in the sector – large language models (LLMs) and GenAI.LLMs have enriched models with access to unstructured text and the ability to distil vast data sets, while GenAI has given rise to pre-trained models able to input data and create, improve, and iterate formulas.“You can also teach an agent what actions to take with reinforcement learning – a process where the agent is given access to an environment/system/programme, and then through trial and error, it repeats the action until the correct outcome is achieved,” Roberts explains.Many hands make light workAgentic AI could have the potential to deliver the level
Mar. Feb 11th, 2025
