methodology
Methodology
The v1 separates the model, collected anonymous signals, and conclusions that require a larger sample.
Report
State of AI Waste 2026 v1
A transparent report page that explains the Four Leaks model, current diagnostic signal collection, and what the first aggregate dataset will publish.
The first public cut of Harmondale’s AI waste model.
Details
Methodology
methodology v1
This page publishes the model and aggregation format. Distributions will be replaced by a larger anonymous sample before being presented as statistical findings.
No lead contact data in the aggregate
3
Example signals
60/100
Average index
67%
High-risk share
four-leaks
Four Leaks of AI ROI
Spend, adoption, leaks, and role drift structure every diagnostic.
next-cut
Next data cut
The report will be updated when the anonymous sample is large enough to publish useful distributions.
FAQ
Is the report already statistical?
No. This v1 is transparent: it publishes the model and prepares aggregation.
What data is used?
Anonymous diagnostic signals: scores, symptoms, team size, and dominant domain.
Report