Structure is key to AI in finance
Artificial intelligence may be the most talked-about force in finance right now, but behind the buzz lies a more complicated reality. In a recent article, Sebastian Reinhard of Caleo argues that large language models (LLMs) promise transformative gains in financial planning yet without structure, governance and domain expertise, they risk delivering more noise than value.
Unlike traditional software, which operates within fixed interfaces and rules, generative models are flexible to the point of unpredictability. That is not to say that they can’t be a powerful force for planning of all kinds, but the article calls for a number of tools. To get the most out of LLMs and turn potential into performance, organisations must embed AI within goal-oriented workflows (also known as “agent mode”) where models understand the working context. Complementing this, the emerging Model Context Protocol (MCP) standardises how AI applications securely access corporate knowledge, connecting to corporate data. Real-world use demonstrates that when LLMs are anchored in high-quality structured data, robust data architectures and governance frameworks, error rates fall.
From our perspective, AI’s effectiveness can be unlocked not only by use protocols, but also by the data that the tool is using, specifically: standardised, machine-readable data. Structured financial disclosures provide precisely the semantic clarity and interoperability that AI systems require. In other words, the future of AI-driven finance will not be built on hype, but on disciplined data foundations.
Read the article here.
