Description
The Financial Reporting Council (FRC) regulates auditors, accountants and actuaries and sets the UK's Corporate Governance and Stewardship Codes. As the Competent Authority for audit in the UK, we set auditing and ethical standards and monitor and enforce audit quality.
The FRC is seeking to strengthen its understanding of how structured data performs in an AI-driven reporting environment. This aligns with the FRC's wider technology and digital reporting agenda, including work to understand the opportunities and risks arising from AI and to support high-quality digital reporting.
The FRC is seeking to commission research services to benchmark the value of XBRL in an AI-driven reporting environment. The study will assess how structured data, including XBRL, performs in comparison with equivalent disclosures presented in PDF and HTML formats.
The research should use clearly evidenced measures, including accuracy, completeness, consistency and explainability, to compare how AI models interpret, extract and reason from disclosures in different formats. The findings will support the FRC's work on digital reporting, structured data quality and the usability of corporate reporting information by preparers, users, regulators and technology providers.
This evidence gap is central to the proposed study. The FRC requires robust comparative evidence on AI performance across equivalent disclosures presented in XBRL, PDF and HTML formats.
The FRC requires the appointment of a suitably qualified external research provider to design and deliver a benchmarking study comparing AI model performance across equivalent corporate reporting disclosures presented in XBRL, PDF and HTML formats.
The benchmarking should include determining which format is the most effective relating to:
• Numerical information,
• Accounting standard-related information,
• Narrative information, and
• Consistency and overall quality of information.
The external provider will be expected to propose a robust research design, identify and justify suitable AI models and testing methods, and produce clear, evidence-based findings.
PLEASE SEE INVITATION TO TENDER FOR FULL DETAILS.