Which data drove this decision?
Facts and judgments link back to the source revisions they used.
Answer “why did this happen?” from one record: the source data, the decision, who approved it and the evidence of the result.
The record is built as the work runs, not pieced together afterwards. Tamper-evident entries stay tied to the versions in force at the time.
For connected tool calls, see what was asked, what was decided and what the target system reported.
Sources, calculations, approvals and results sit in one connected case history.
Inspect the recorded evidence behind your native agents’ governed actions.
Answer an employee’s, customer’s, vendor’s or partner’s “why?” with the decisions and source records behind their case.
Connect the data to the rules, authority, exact approval and recorded outcome. When the evidence is missing, the gap stays part of the explanation.
Facts and judgments link back to the source revisions they used.
The exact versions and the approved proposal are on the record.
A confirmed change, an attempt and a missing confirmation stay distinct.
A source-to-field map explains data movement. A business outcome also involves identity, policy, judgment, approval and execution.
The published application version and the individual run connect the sources, decisions and results. Each relationship keeps its meaning across systems and waits.
| Concept | The question it answers | What supports the answer |
|---|---|---|
| Provenance | Where did this artifact originate? | Source identity, revision, derivation and generator. |
| Data lineage | How did information move or change? | Field, document, dataset and feature transformations. |
| Context provenance | What was supplied to this decision? | The input records, calculated facts and versions supplied to the decision. |
| Decision lineage | Which judgments formed the result? | Model, deterministic, routing, policy and human decision records. |
| Execution lineage | How did this run reach its outcome? | Inputs, authority, state, decisions, attempts and external effects. |
| How records link | How exactly are two records related? | Explicit links such as derived from, constrained by and authorized by. |
| Evidence | Which retained record supports the claim? | Receipts, snapshots, version references, digests and protected source records. |
Every link says what it means, not just when it happened. “Approval authorized proposal r5” states which decision permitted which action.
Follow INV-1042 through eight stages, and switch cases to see what the record shows when something goes wrong.
Inspect the sources, versions and relationships behind a confirmed payment, missing confirmation, changed proposal or reconciled no-payment result.
The ERP receipt confirms payment PAY-4821 for the approved invoice r5.
The declared boundaries have supporting records for this path.
Sending a payment request is not proof it was paid. The change is marked confirmed only once your ERP confirms it.
A missing confirmation leaves the payment outcome unresolved. The record shows the attempt and the gap, while a superseded approval stays tied to its original proposal.
Application, proposal, source, policy and approval versions stay attached to the run through durable waits. You can see what governed the recorded decision, even after the application changes.
See how work survives waitsYou can see whether a tidy timeline is a complete one.
The Lineage Coverage Contract defines the records each consequential path needs: sources, decision inputs, policy, approvals, attempts and a confirmed outcome or a reconciled no-change result.
Compile-on-Publish checks required evidence domains and links on the relevant workflow nodes. The Traceability Map connects runtime behavior to the published application version.
At runtime, missing expected evidence stays visible too.Each required record shows as present, missing or not needed for that path. Sources, decisions, approvals and outcomes stay connected to the same run.
Coverage follows the path actually taken.The record shows available content, protected references, versions, retention state and integrity information together.
Content and access are shown in their current state.Selective snapshots, protected references, source revisions, redacted excerpts and confirmation metadata keep the explanation connected. Authoritative content can remain in the system that owns it.
Tamper-evident records let you verify whether the retained evidence has changed.
Retention rules and classifications stay on the record, along with any known gaps.
Evidence is protected like the data it describes.
Tenant, purpose, business scope, classification and source permissions determine what each reader can inspect. The six views below connect those permissions to the questions each team needs to answer.
State, waits, attempts, retries, timing and reported effects.
Graph nodes, schemas, mappings, dependency versions and errors.
Identity, delegation, assurance, scope, policy and capability eligibility.
Material evidence, judgments, obligations, escalation and consequence.
Versioned records, approval binding, effect evidence and known coverage gaps.
Relevant business state, the human decision and the resulting outcome.
Each view shows only what that person is currently allowed to see.
Frozion stores or references each record and its links, connecting the systems you already run around the same application run.
Explore the AI Data PlaneUse the connected record to inspect a run, compare outcomes and reconstruct the history. Replay uses that record under its own access controls.
Explore Behavior IntelligenceData catalogs, identity systems, model gateways, workflow engines and observability tools retain their own records. Business systems retain committed state. Frozion connects those references to the decision, approval and result in the business case.
Give auditors scoped, read-only access, or export the cases they sample. The record keeps source restrictions, retention state and known gaps visible.
Every recorded link is explicit, and every known gap is visible.
Bring one business process. We’ll show the record it would leave, from source data to the confirmed result—or the gap that still needs resolving.