Vannus Blog
Long-form thinking on AI vendor jurisdiction, data sovereignty, and procurement due diligence.
We re-checked 25 of our records. Here is what we could check, and what we could not.
25 records drawn at random. A script may re-read 7 of them: every quote was still on the page, and one record stated a country its document does not, which we corrected. Why the other 18 cannot be re-read by script.
We used to grade AI vendors. Here is why we stopped.
Until 8 September 2026 Vannus published a letter grade, a tier and a score. What they were, why they went, what replaced them, and every place they lingered after that date, with the day each came off.
One vendor, four contracting entities — and we had recorded one.
Neo4j’s own agreement names four possible contracting entities, selected by where you are incorporated and how you bought. Our record said one. A third of the vendors we have read closely have a split like it.
Historical: we added 69 AI tools, and held them back twice before publishing.
Two attempts would have published failing grades on real companies because of gaps in our own database — a 97% failure rate the first time, and an F on the most privacy-respecting vendor in the batch the second. What went wrong, and the three changes it forced in the methodology. Kept as published, with corrections; the grades it describes were withdrawn on 8 September 2026.
Historical: we re-graded every AI tool in our catalog. 233 of 287 grades changed.
In July our heaviest scoring criterion was found to be measuring the wrong thing — it credited tools by name, not by what they run, and scored resellers above builders. We fixed it and re-graded all 287. Grading itself was withdrawn on 8 September 2026; the evidence rule outlived it: a finding about a vendor's model traces to the vendor's own documentation, or says it could not.
The Shadow AI Audit — a 90-minute checklist for ops leaders
A four-step inventory you can run in one calendar block. Pulls AP data, surfaces shadow accounts, marks what you cannot yet establish about each vendor, and ends with three written decisions. No SaaS-management platform required. Part 1 of the AI Tool Consolidation for SMBs series.