EN ES
Contact

Report

We scored twenty real AI purchases. The median: 28% of the standard

Enpath · · 22 pages

Hands reviewing a document beside a calculator.

Key takeaways

  • Median 28%; maximum 57%; half the processes below 30%.
  • Operation is no man's land (11.7%): seven checks met out of sixty, zero published metrics, a single verifiable audit.
  • The four generative-tool processes score worst of all (median 19%) — they arrived through minor channels, with no documented problem.
  • The four structural zeros are fixed by instruction, without touching a law.
  • The baseline now exists: any AI procurement policy the region adopts can be measured against the 28%.

The region's first empirical measurement of public AI procurement: twenty processes from five countries against thirty checks. 27.6% aggregate compliance and four absolute zeros — none of which needs a law to fix.

The full document is currently published in Spanish; an English edition is prepared when demand warrants it. The Spanish record carries the complete summary and contents.