? For financial professionals, the latest advances in GenAI do not offer the same added value as for IT, legal and media experts. Why?
Because while large Gen-AI models are excellent at handling text, images and sound, financial data is mainly number-based and does not have a standardized syntax like program code. Each company has its own cost and revenue structure that requires individual logic and presentation.
The MS365 Copilot for Excel clearly demonstrates this challenge: when dealing with data, it is not as effective as its counterparts for emails, documents or meetings.
?? A recent and fascinating paper from Microsoft introduces the “SpreadsheetLLM”, which converts spreadsheet content into a new format that LLMs can work better with. The aim is to answer questions about Excel spreadsheets, manipulate data and even create new spreadsheets using natural language input.
?? The startup Formula Insight takes a similarly interesting approach. They are building an API LLM interface that lists each cell of the Excel spreadsheet in a JSON file with all its properties. In this way, the spreadsheet becomes searchable for LLM.
The range of tools currently available to make the day-to-day work of financial experts easier is still very limited.
?? We at DocuTrust have recognized precisely this problem and are working on a technology that understands financial data and the underlying relationships.
??? “Financial Table Understanding” is one of the core problems we are working on. Stay tuned for the innovations we will be presenting to you soon!
