Could XBRL tame the comment letter deluge?
When a single rule proposal draws more than 181,000 comment letters, how does anyone make sense of them? In a new Point of View blog, Campbell Pryde, President and CEO of XBRL US, argues that structured, machine-readable comment templates would let regulators and market participants put AI to better use.
That 181,000 figure comes from an AI-powered tracker of responses to the Securities and Exchange Commission’s (SEC) proposal on semiannual reporting. US agencies publish thousands of final rules each year, each typically preceded by a lengthy proposal and a public comment period, and almost all comments arrive as unstructured PDFs. AI can parse them, but only by inferring structure, which costs time and energy and produces inconsistent results across different large language models (LLMs). Pryde also notes that the SEC’s own AI use case report lists 49 use cases across 22 divisions, with several teams separately building tools for comment-letter review.
His proposal is a structured questionnaire producing data natively in XBRL, with each response tied to a clearly defined tag, a Legal Entity Identifier (LEI) and dimensions such as commenter type. Regulators could then query every letter at once, and law firms, researchers, journalists and investors could use the same data to track sentiment on proposed rules. Building on an established standard also means drawing on thousands of existing tools, where a custom schema would need its own from scratch.
The idea lands at an interesting moment. In his speech reported elsewhere in this newsletter, SEC Chief Economist Joshua White told the Investment Company Institute this month that the comment process is one of the ways the Commission learns, and asked commenters for well-evidenced economic data. Structuring that input would make it much easier to learn from at scale.
Read the full blog here.
