Retention classes
Records retained by class, contract, and lawful requirement — with defaults for governance-grade evidence.
HeldAuditability for governed agentic execution
Make every AI workflow traceable, reviewable, and evidence-backed.
ZoikoVertex records the actions, decisions, approvals, identities, evidence, exports, and exceptions behind agentic workflows — so enterprise teams can move faster without losing accountability.
Five connected evidence surfaces + legal holds
Audit TrailDecision LedgerEvidence VaultIdentity LedgerForensic HubLegal Holds
Why auditability matters
Unaudited AI is an unanswered question.
When agents act at machine speed, “trust us” isn't an answer for legal, security, or the board. Auditability lets your organization answer, with proof:
Q1What happened?
Q2Who acted?
Q3Why was it decided?
Q4What proof exists?
Q5What changed?
Q6What was approved?
Q7What was blocked?
Q8How is it reviewed?
Five-surface evidence model
A system of connected records — not a flat log.
Auditability is five linked surfaces. Each answers a distinct question, and every event threads through them.
Surface 01
Audit Trail
What happened?
A chronological record of human, AI, workflow, policy, integration, and system events.
Surface 02
Decision Ledger
Why was it decided?
A structured record of rationale, approvals, policy basis, and human judgment.
Surface 03
Evidence Vault
What proof exists?
Sealed packages of prompts, outputs, approvals, published content, policies, and manifests.
Surface 04
Forensic Hub
What can be reconstructed?
Case-based reconstruction of flagged, disputed, escalated, or high-risk events.
Surface 05
Identity Ledger
Who acted, with what authority?
Role, session, permission, MFA, and privileged-action records linked to audit events.
What gets recorded
Only what makes an action defensible.
Governance-grade records capture enough to prove accountability — without over-collecting. Retention is set by class, contract, and lawful requirement.
| Data class | What is stored | Retention position |
|---|---|---|
| Governance audit events | Event ID, tenant, actor, action, object, timestamp, status, risk, policy, evidence links, hash references | 7-year default for governance-grade records |
| Decision records | Decision ID, rationale reference, approver, policy basis, review stage, timestamp | 7 years or match contract / legal hold |
| Evidence records | Prompt/output snapshots, approvals, final content, export manifest, package hash, retention class | 7 years where evidence-linked; shorter for drafts |
| Identity / access records | Role at time of action, MFA state, session, privileged access, permission changes | 2–7 years by risk and privilege |
| Forensic case records | Case timeline, source event, evidence bundle, investigator notes, outcome | 7 years after closure or legal hold |
| Export & access records | Export reason, exported by, recipient/destination, manifest hash, access history | 7 years for audit-sensitive exports |
Audit journey
Watch one event become defensible proof.
Step through a real governed workflow event. Each step activates a surface, links a record, and stamps a status — ending in an exportable, verifiable bundle.
- 1Event occursAn AI agent generates, edits, approves, publishes, blocks, exports, or escalates content.
- 2Decision capturedA decision badge appears when approval or policy judgment exists.
- 3Evidence sealedThe evidence package is stored and sealed.
- 4Actor verifiedThe identity badge shows role, session, and authority.
- 5Case openedconditionalOptional — escalate to a Forensic Hub case if flagged or disputed.
- 6Bundle exportedAn audit bundle is generated with a manifest; the export is itself logged.
Audit record · liveZV-EV-8842190
Logged
You see: an agent generated a regulated Q3 claim and it entered the workflow.
Audit Trail
- Actor
- Agent · Campaign Copywriter
- Action
- Draft generated
- Object
- Q3 Regulated Claim / #1204
- Timestamp
- 2026-07-08 14:22:07 UTC
Decision Ledger
- Approver
- J. Reyes · Clinical Reviewer
- Policy basis
- Regulated Claims Policy v4
- Review stage
- Approved with edits
Evidence Vault
- Package
- EVP-1204-A
- Contents
- Prompt · Output · Approval · Policy
- Retention class
- Governance · 7yr
Identity Ledger
- Role at action
- Reviewer
- MFA state
- Passed
- Privileged access
- No
Forensic Hub
- Case
- CASE-0731 · optional
- Source event
- ZV-EV-8842190
- Status
- Under review
Step 1 / 6Next step
Enterprise controls
Records that legal and security can rely on.
The controls procurement asks about — retention, holds, access, redaction, export, and tamper evidence — built in, not bolted on.
Legal holds
Preserve evidence during disputes, investigations, and regulatory requests; suspend deletion where required.
On hold
Role-based access
Scope who can view, export, and act on audit records across tenants and workspaces.
◆ Role-bound
Redaction
Protect sensitive fields in exports and reviews without breaking the integrity of the record.
Redacted
Export manifests
Controlled bundles with reason, recipient, manifest, and hash — every export logged as a new audit event.
Exported
Tamper-evident verification
Hash references let reviewers verify that an evidence package has not been altered since sealing.
▤ SealedWhere auditability proves itself
Every material action, evidenced.
Approval evidence
Prove that regulated or brand-sensitive content was reviewed and approved before publishing.
Evidence · approval + policy basisAI agent actions
Trace what each agent generated, edited, or attempted — bound to role and authority.
Evidence · actor + action recordPolicy blocks
Show what was stopped, by which policy, and why — not just what shipped.
Evidence · blocked-action recordPublishing proof
Preserve the final published artifact alongside its approvals and manifest.
Evidence · sealed packageDisputes & investigations
Reconstruct a disputed event end to end from source event to evidence bundle.
Evidence · forensic casePrivileged activity
Log privileged access, identity changes, and exports for security review.
Evidence · identity + access logWhere auditability fits
Proof, policy, and principles — three distinct layers.
Auditability is the proof and traceability layer. It works alongside — not instead of — your policy framework and AI principles.
Auditability
The proof engine: what happened, who acted, why, what evidence exists, and how it's reviewed and exported.
Answers · “Can you prove it?”Compliance & Governance
The policy and control framework: the rules, roles, and processes that govern how AI execution is allowed to happen.
Answers · “What are the rules?”Responsible AI
The principles: bounded, traceable, reviewable AI actions, linked to decisions and supported by evidence.
Answers · “Is it done responsibly?”Prove it to your committee
Bring auditability into your evaluation.
Auditability demo
See the five surfaces, evidence bundles, and export manifests live — mapped to your workflows.
Request an Auditability Demo
Auditability brief
A committee-ready overview of records, retention, holds, and export controls.
Download Auditability Brief
ROI & Governance Audit
Quantify faster reviews and lower governance friction alongside the evidence model.
Run ROI & Governance AuditAuditability FAQ
Answers for legal, security & procurement.
ZoikoVertex auditability
The safest commercial path to agentic AI execution.
Every material action traceable, every decision linkable, every actor bound to authority, every evidence package reviewable, every export defensible.