The record is the product.

The evidence is the raw record

Every payload a platform sends us is kept exactly as it arrived — content-hashed, written to the ledger, and stored unmodified with its provenance: where it came from, how it was recognized, which decoder version read it. Capture never waits and never drops. If a format isn’t recognized yet, the payload is kept anyway and decoded when it is.

Chain of custody, reproducible derivation

What you read in the app — conversations, findings, dollar figures — is a derived view over that raw evidence, and it can be re-derived at any time. When a decoding improves, the record isn’t rewritten: the new version is derived alongside, and anything superseded is marked superseded, on the ledger. A bad decoding can never touch the evidence itself.

Your rules, versioned like contracts

The rulebook your conversations are judged against is stored as immutable revisions — every edit creates a new revision, and every finding cites the exact revision it was judged under. You can always answer “what rule was in force when this happened?”

The append-only Trust Ledger

Every significant event — a conversation ingested, a finding raised, an action you took, a rule you changed, consent you recorded — lands in an append-only ledger in the same database transaction as the change itself. The ledger never gets edited. It is your business’s own evidence trail: if a customer, vendor, or regulator ever asks what your AI actually did, the record is yours — not your AI vendor’s.

Independence, graded on every finding

The less a platform can edit what we see, the stronger the evidence. Every agent shows its capture grade in plain words — Tier A, independent capture, recorded by Anzuna directly; Tier B, platform-signed events, verified on arrival; Tier C, reported by the agent’s own stack, and graded as such — and every finding carries its source’s grade on the finding itself. The coverage map states every platform’s path, so you always know exactly what’s watched.

Independent by structure, not by policy

Anzuna doesn’t build agents, doesn’t sell agents, and isn’t owned by anyone who does. We work for the business being represented — no one else. Industry watchdogs keep getting absorbed by the platforms they watch, and a policy promise survives exactly until the acquisition that tests it. Here the independence is structural: our only product is the record, and our only customer is you.

Estimates marked as estimates

Dollar figures carry a confidence class, and anything estimated is shown with a tilde (~) — never fake precision. Findings the judge is unsure about are tagged for your review instead of being quietly dropped or quietly asserted.

What we never do

• We never operate, modify, or speak for your agents — we only watch and report.
• We never edit the ledger or the raw evidence. Corrections are new entries, never rewrites.
• We never assume coverage. Your coverage map states every platform's exact path, and the shortest route to strengthening it.
• We never start monitoring before you've confirmed consent — the consent gate is recorded on the ledger, and consent-notice templates are provided in the app.