When Old Tech Beats New Tech: MSCI Institute on XBRL for Sustainability Reporting

About the source

This is an MSCI Institute insight piece, not our research. It was published on 30 September 2025 by Linda-Eling Lee (founding director of the MSCI Institute) and John Turner (CEO of XBRL International). We summarise it here because it directly informs the European debate on mandatory XBRL tagging for sustainability reports under ESRS / CSRD.

Read the original: When old tech beats new tech: Sustainability reporting with XBRL.

The core claim

Financial reporting already runs on XBRL in major markets. Sustainability reporting mostly does not: disclosures still land in PDFs with no consistent machine-readable layer. MSCI Institute argues that adopting XBRL for sustainability would let investors extract material information up to 10 times faster, on average, and at a fraction of the cost of using AI to pull unstructured data from PDFs — based on their analysis of India’s XBRL-based sustainability filings.

AI can help with unstructured documents. The authors’ point is narrower: on speed and accuracy for tagged fields, structured digital formats still beat PDF archaeology.

Where digital sustainability reporting stands

According to the piece:

XBRL is mainstream for financial reporting in jurisdictions covering companies that together represent a large share of global market value.
For sustainability reporting, India is the only country so far to mandate XBRL (under BRSR). South Africa allows voluntary XBRL sustainability filings.
ISSB and European ESRS taxonomies already exist for digital reporting, but regulators have been slow to require them.
In Europe, the Commission has so far not instituted XBRL-formatted disclosure under CSRD while considering Omnibus simplification proposals.
Broader pattern: many jurisdictions roll out sustainability reporting rules without matching digital-reporting capabilities.

What their India comparison finds

MSCI Institute compares AI-driven PDF extraction with XBRL parsing of Indian BRSR digital filings. Early findings they report:

	PDF + AI	XBRL
One-time setup	4 days (AI prompting)	10 days
Subsequent datapoint onboarding	5 days	0.5 days
Example extraction cost (1,000 companies / year)	USD 500 for LLM API use	No cost (tag-based scripts)
Reliability for tagged fields	Variable / probabilistic	High / deterministic for correctly tagged values

So the one-time digital onboarding can take longer than a first AI pipeline. Once in place, each new datapoint is far cheaper and faster to process. Accuracy for XBRL is described as high for tagged values, with remaining errors mainly from human tagging mistakes (wrong financial-year attribution, blanks represented as zeros), not from model hallucination.

The authors are explicit that the two approaches are not mutually exclusive. Their preferred framing: use XBRL tagging as the foundation and apply AI on top for analysis — not treat AI PDF extraction as a reason to skip digital tagging.

Why this matters for the EU debate

Europe already knows XBRL from financial reporting (including ESEF). The open question for sustainability is whether mandatory machine-readable tagging under ESRS should proceed, or whether claims that “AI can scrape the PDFs” should weaken that requirement.

MSCI Institute’s public analysis lands on the side of digital-first disclosure: structured data at the source reduces repeated extraction cost for every investor, regulator, and civil-society user. That aligns with the engineering case we make in our Call for Evidence on mandatory XBRL tagging for ESRS vs AI claims and the short opinion Is XBRL Tagging for Sustainability Reports “Archaic Nonsense” — or a Chance for Digital Reporting and AI?.

Access the source

Lee, L.-E., & Turner, J. (2025, September 30). When old tech beats new tech: Sustainability reporting with XBRL. MSCI Institute.