Contact
Evaluate DSI Audit, or tell us where it is wrong.
DSI Audit is at v0.2.1 — verified to build and run from a clean install — offered for evaluation and pilot. The most useful thing you can do is run it locally and challenge it — the install is a short local setup, the numbers are reproducible, and we state the limitations up front.
Who this is for
- Teams building or deploying advisory AI who want to see which expected option paths their assistants keep in view — and evidence they can reproduce.
- AI-governance and assurance reviewers who need decision-space evidence to attach to a review (DSI Audit supports documentation; it does not certify compliance).
- Researchers working on evaluation methodology who want to scrutinise the method and the numbers.
How to start
Run it locally
Follow the evaluation guide to a working audit in a short local session.
Talk to us about a pilot
For a supplied-response audit alongside your own stack.
Send objections
We'd rather hear what doesn't convince you.
We are particularly interested in evaluation or design-partner work where a model, prompt, retrieval or orchestration change may have silently removed configured information that an existing assurance process already cares about.
Email enquire@decisionspaceintegrity.com. The product repository is private while evaluation access is managed directly. The public replication repository remains available for research review.
About
Built by Andrew J. Cousins, an enterprise technology specialist with more than two decades of experience in large-scale banking systems, combined with research into decision-space collapse and AI assurance.