Methodological note on the use of artificial intelligence: “The Last Mile of ESG Data”

This work examines how reliable ESG data is at the source and what role artificial intelligence plays in collecting, checking and consolidating it. It is therefore appropriate to state how a generative AI assistant (Claude, by Anthropic) was used in preparing it.

Uses

Drafting: the article was written from my own research notes and an earlier draft in French. The assistant helped shorten it, adapt it into English and test different structures and tones. [I reviewed, edited and approved the final text.]
Visuals: the two figures (the ESG data chain and the Indirect/Direct impact boxes) were produced with the assistant’s help, based on content I approved.
Source checking: the assistant helped find and read sources (for example Shen et al., 2025, and the IEA’s data on data-centre electricity use). Each claim was checked against the source before publication.
Publishing: practical help with formatting and publishing the article on WordPress, including SEO settings and image alt text.

What AI did not do
It produced no interview data and no quotations. The practitioner observations in the article come from my own interviews. [Interviewees are not named.]
The argument, the choice of sources and the conclusions are mine.

Safeguards
Every factual or numerical claim was checked against the primary source. Several were corrected or removed in the process, for example a statistic whose source did not support the original wording.