Evidence Vault

Prove how AI decisions were made

A decision is easier to defend when the evidence is created at the same time as the decision.

Evidence Vault adds a sealed decision record to Decision Deck so every AI-driven decision carries proof of what was known, what policy ran and why the final action followed.

Premise

That gives teams two things they need when automating high-impact decisions

At the point of action

Evidence at the point of action.

Creates the decision record as the decision runs, before review, audit or challenge begins.

After action

Proof after action.

Shows which evidence was used, which policy applied and where human judgement changed the path.

Approach

From scattered records to decision evidence

01 · Today Fragments across systems

Most organisations can find fragments of a decision somewhere. Evidence may sit in source systems, logs, documents, workflow notes, model outputs or reviewer comments. The problem starts when someone asks why a specific decision happened.

02 · With Evidence Vault One decision record

Evidence Vault turns those fragments into one decision record. As each decision runs, it captures the evidence, policy checks, version context, actions considered, actions held back and final decision state.

Stop reconstructing decisions after the event.
Controls

Evidence Vault controls

01

Decision record

Create one structured evidence record for each decision.

02

Policy trace

Show which policy version ran and what it allowed, blocked or routed.

03

Evidence use

Record what evidence was used, missing or rejected.

04

Version context

Preserve the policy, model and workflow versions active at the time.

05

Human judgement

Capture review, approval, escalation and override rationale.

06

Verification

Seal the record so later change is visible.

When to use

When to use Evidence Vault

Use Evidence Vault where a decision may need to be inspected, defended or improved later. It is most valuable when the cost of explanation is high and the evidence behind the decision cannot be left to logs, notes or memory.

Evidence Vault is particularly useful when AI-driven decisions are governed by policy-as-code, human review or model versions that change over time. It gives teams a decision-level record of what actually happened, using the context active at the time.

Where it applies

Where Evidence Vault shines

Challenged decisionsAnswer why a decision happened without rebuilding the case from disconnected systems.
Policy proofShow which policy ran, which checks passed and which actions were blocked.
Human reviewRecord where judgement entered the decision and why the reviewer changed or confirmed the path.
Versioned decisionsInspect historic decisions against the policy, model and workflow versions active at the time.
Governance reportingBuild reporting from evidenced decisions, with case-level proof behind the aggregate view.
AI assuranceGive teams stronger evidence for how AI-supported decisions behaved in production.
Operating proof

Decision evidence as operating proof

Evidence Vault turns decision evidence into part of the operating method. The decision does not produce an outcome first and an explanation later. The same process that produces the outcome also produces the record. That makes each decision easier to inspect, easier to defend and easier to learn from.

Evidence Policy Versions Review
How it works

How Evidence Vault works

01

Evidence requirement

Define what must be captured for the decision to be reviewed later.

02

Runtime capture

Record the evidence, policy checks, model outputs and review activity as the decision runs.

03

Sealed envelope

Preserve the decision record with the version context active at the time.

04

Replay and learning

Inspect individual decisions, compare later versions and learn which evidence, policies and review paths improve outcomes.

Impact

Business impact

01

Risk and compliance

Give teams clearer proof of how policy, evidence and review shaped each decision.

02

Responsible AI

Show how AI-supported decisions operated, including uncertainty, held actions and human involvement.

03

Operations

Reduce the effort spent reconstructing decisions for complaints, appeals, reviews and audits.

04

Product and technology

Compare decision behaviour across versions and improve the evidence captured by each workflow.

Questions

Common questions

How is Evidence Vault different from logging?+

Logs show system activity. Evidence Vault shows how the decision was made. It records the evidence, policy checks, versions, actions and review activity that shaped the outcome.

What is a Decision Evidence Envelope?+

A Decision Evidence Envelope is the sealed record for one decision. It captures the material evidence, policy version, model version, action path, human review and final decision state active when the decision was made.

How does Evidence Vault work with policy-as-code?+

Policy-as-code controls what the system is allowed to do at runtime. Evidence Vault records which policy ran, which checks passed or failed and how those checks changed the decision path.

Does Evidence Vault prove compliance?+

No. Evidence Vault does not replace legal, compliance or governance teams. It gives them stronger evidence of what happened, what was checked and why the decision followed.

Does Evidence Vault record actions not taken?+

Yes, where they mattered to the decision. The record can show when an action was held back because policy, evidence, risk or customer treatment controls changed the route.

Where does human judgement fit?+

Human judgement remains part of the governed decision process. Evidence Vault records the review route, evidence shown, reviewer action, override rationale and final outcome.

Make every AI decision easier to defend

Evidence Vault brings decision evidence directly into AI-driven workflows. Use it where the ability to prove what happened matters.