SRC Digital Insurance Services, led by XBRL proponents Alfred Berkeley and K. Dixon Wright, has launched a research initiative to develop and evaluate AI tools for converting Department of Transportation (DOT) “Application for Payment” PDF reports into XBRL data.
The program aims to test how different AI approaches interpret the same DOT source documents and measure the impact of inconsistencies and hallucinations on data integrity. The initiative will fund $5,000 seed contracts to teams developing AI models for extracting and tagging payment application data into an XBRL taxonomy. Each project is required to produce a converted XBRL dataset from publicly available DOT payment application PDFs, an analysis comparing results across AI tools to identify divergences and hallucination risks, a catalog of data elements used by DOTs and capital markets, and a stakeholder review and comment process to inform recommended XBRL taxonomy elements across public and private sectors.
The research seeks to demonstrate whether varying AI interpretations of identical DOT documents produce materially different outputs and to quantify how those differences affect the usefulness of the resulting data for underwriting, project monitoring, and capital markets access. Findings from this work are intended to guide taxonomy design, improve data interoperability, and support more reliable, actionable datasets for public agencies and marketplace participants.
K. Dixon Wright commented on the initiative, stating, “Digitizing isn’t just moving spreadsheets to the cloud. A cloud full of PDFs and spreadsheets does not produce actionable data; AI interpretations can introduce hallucinations that undermine data value.” Alfred Berkeley added, “XBRL provides high-integrity data in an internationally recognized standard. This work supports interoperability requirements and helps make data genuinely actionable for decision makers.”
The first seed contract has been awarded to Carson Whittiker, who is receiving mentorship from Professor Jorge Vanegas of the Texas A&M College of Architecture. Professor Vanegas noted, “Transportation and AI are major focuses at Texas A&M. These seed contracts help students and entrepreneurs deliver solutions that combine mandated reporting requirements with operational gains for stakeholders.” Carson Whittiker, a student from Professor Vanegas’s entrepreneurship class, stated, “At IntraData Exchange we will convert DOT PDFs into XBRL that capital markets and stakeholders can access to improve digital underwriting and project monitoring, expanding financing access for small and local businesses.”
Additional seed contracts will be offered through partners including the Johns Hopkins Technology Venture Fund, where Alfred Berkeley is active. The research and stakeholder review period will engage trade associations, public agencies, and industry groups to help define which data elements should be included in a sectoral XBRL taxonomy.
