Seed funding for AI projects to convert PDFs into XBRL
SRC Digital Insurance Services has launched a research initiative to explore whether artificial intelligence can reliably convert US Department of Transportation (DOT) “Application for Payment” PDF reports into XBRL data. The project offers 5,000 dollar seed contracts to entrepreneur–student teams, working with university mentors, to trial different AI models on the same documents and measure how consistent their outputs prove to be.
Each project will deliver an XBRL dataset, an analysis comparing outputs from competing AI tools, and a stakeholder review to inform proposed taxonomy elements. The aim is to see whether varying AI interpretations of the same DOT source documents produce materially different data, and crucially, to assess what this means for trust, comparability and the usefulness of the resulting information.
What stands out here is the explicit link between AI and taxonomy development. As Alfred Berkeley of SRC noted, “XBRL provides high-integrity data in an internationally recognised standard.” If AI can be harnessed to inform taxonomy design, this could help bridge the messy world of PDFs and the structured world of digital reporting. What is less clear is the ultimate direction: will this lead to permanent AI conversion pipelines, or serve as a stepping stone while reporting entities adopt structured digital reporting frameworks?
The first award has gone to IntraData Exchange, led by student entrepreneur Carson Whittiker with mentorship from Professor Jorge Vanegas at Texas A&M. Further seed contracts are expected, in partnership with Johns Hopkins and other institutions. We’ll be following the results with interest.
Read more here.
