Platform
Pilot scope · mortgage-tenant-1 · internal policy · pre-production evidence

Capabilities of the governance engine.

A practical view of what TraceLogic provides under the surface. Each capability is part of the same governed decision lifecycle, captured into the same evidence record.

Twelve capabilities, one record

Each capability writes into the same governed decision artifact. There is one source of truth per decision, not twelve scattered logs.

📥
AI-Assisted Intake
Documents and case inputs are captured and structured. Field provenance is preserved alongside extracted values, so reviewers can trace each value back to its source.
Evidence: Operational Smoke (P0-5)
📐
Decision Checks
Decisions are evaluated against configured rules. The check is deterministic, repeatable, and produces an output that is recorded in full alongside the inputs that drove it.
Evidence: Operational Smoke (P0-5)
🗂️
Policy Trace
Each decision carries the policy version it was evaluated against and the specific rule path that produced the outcome. Policy logic is visible as part of the evidence, not buried in code.
Evidence: Operational Smoke (P0-5)
👥
Human Review
A reviewer sees the decision output, the supporting evidence, and the policy trace in one view. They approve or return the case based on what is presented, not on a separate spreadsheet.
Evidence: Operational Smoke (P0-5)
🛡️
Controlled Override
A governed exception path. Override requires a reason code from a closed vocabulary, a written rationale, a supporting evidence reference, a risk acknowledgement, and Separation of Duties (the submitter cannot approve the override on the same case). Every override is sealed into the decision record and is visible during audit replay.
Evidence: Operational Smoke (P0-5)
🔒
Frozen Artifact
On approval, the case is sealed into a decision artifact: evidence, policy result, decision output, review status, and governance metadata, recorded immutably.
Evidence: Operational Smoke (P0-5)
🛂
Execution Gate
Execution is authorised against the approved artifact. Direct action is not the trigger. Approval, attestation, token-based handoff, and role separation are all part of the gate.
Evidence: Live Atomic Consume (P0-6)
Replay
Decisions can be replayed later from the stored artifact only. No live recomputation. No reliance on memory or current state. The frozen evidence is the source of truth.
Evidence: Operational Smoke (P0-5)
📊
Trust Dashboard
A business-readable view across coverage, exceptions, approval lineage, and replay evidence. Designed for managers, compliance, risk, and audit, not engineers.
Evidence: Observation Dashboard (P0-I6a)
🏢
Tenant-Scoped Governance
Configuration, policy, evidence, and replay are scoped per tenant. A single deployment can support multiple decision domains without crossing the controls between them.
Evidence: Operational Smoke (P0-5) — one tenant in pilot
⚖️
Role-Based Lifecycle
Distinct roles for submitter, reviewer, approver, and executor. The lifecycle enforces the separation; permissions are not assumed.
Evidence: Operational Smoke (P0-5)
📦
Evidence Package
Every governed decision can be exported as an evidence package: source documents, intake form, decision result, policy trace, reviewer identity, approval status, and execution record.
Evidence: Assurance Index

What TraceLogic now blocks, proves, and replays

Five hardening workstreams have closed end-to-end: source on origin, running in the pilot, verified by source-layer tests at the deployed commits.

Blocks unsafe execution. TraceLogic blocks unsafe execution. A single execution token can be consumed at most once. Tokens are refused before mint when the framework adoption is not currently activated. Tokens with the wrong shape, missing fields, or framework-binding mismatch are refused at consume time.
Proves framework-bound governance. Every execution token carries a typed, signed claims envelope bound to the framework the decision was made under. Issuance is checked at the gate, not reconstructed later at audit time.
Replays the decision as history. Replay runs against frozen evidence sealed at decision time; the live policy engine is not invoked. The current adoption-registry state is surfaced as informational metadata alongside the historical outcome, without ever changing the historical outcome.
Honest scope. Two specific demonstrations remain deferred: live fail-closed mutation smoke and authenticated replay smoke. Source-layer test coverage at the deployed commits remains the engineering evidence.

The engineering record is at docs/evidence/P0_2_BUYER_RELEASE_NOTE.md.

Built around the governance lifecycle

The platform is structured around the decision lifecycle, not around features. Every capability has a place in the lifecycle, and the lifecycle is the contract between the capability and the user.

Deterministic by default. The same inputs and policy version produce the same decision.
Evidence preserved. Source documents and field provenance are part of the record, not metadata.
Human accountability retained. The platform supports review and approval; it does not replace them.
Replay from frozen evidence. Replay never depends on live systems or current state.
Tenant isolation. Decisions in one tenant cannot affect or be visible to another.
Pilot stage. The platform is currently positioned for controlled pilot discussions. It is not claiming enterprise production deployment, formal client validation, or regulatory certification.

See it on a real decision

The pilot demo walks each capability through a governed mortgage forbearance decision.