World Bank Loans: Health Report

AITS pulled two sets of public records from the World Bank: its statement of loans, and the loan agreements it publishes. Every loan and every agreement was checked against the KPIs below, including how well the two sets agree with each other, and the results were rolled up into one composite Repository Health score.

A dimension's score is the average of its KPIs. The composite is the average of the measured dimensions.

"Not measured" means the source data did not carry what that KPI needs, so it is left out of the score instead of being counted as a failure.

16 actions, largest gain first. Completing all of them raises the composite score from 89.0% to at least 99.7%. AI Readiness is not on the list because it is not fixed directly: it rises as these are completed, so the real result would be higher.

Expected gain is what the composite score rises by if every affected loan or agreement is brought to a pass on that KPI. In an engagement, CTS recommends action on each record, staff confirm or correct, and the score is kept current as records are added.

"Checks passed" is the share of nine agreement-level checks each agreement passed, averaged over the region. "Ready for AI" is the share of agreements that pass every check an AI assistant relies on.

• The sources: the World Bank's statement of IBRD loans (latest monthly snapshot, from its Finances One open data, licensed Creative Commons Attribution 4.0) and the loan agreements published in its Documents & Reports library, both read on October 6, 2026.
• Every record, not a sample: each loan and each agreement was checked as it was loaded, the same way CTS assesses each item as it arrives in your virtual data repository.
• Two sets of records, compared: loans and agreements are matched by project. That is what shows a loan with no agreement on file, or an agreement that disagrees with its loan.
• Nothing was edited: the data was read from the published copies. The World Bank's own systems were not involved.
• What is different in an engagement: you choose the KPIs, the weights and the target; the text of each agreement is assessed as well as its details; Security is measured; and the AI learns from every finding for your staff to confirm or correct. This teaches the AI so that it gets better over time.

Source data: The World Bank, IBRD Statement of Loans and Guarantees (Creative Commons Attribution 4.0) and Documents & Reports. Agreement titles link to the World Bank's own copy.