A trace shows the technical path. Behavior Intelligence shows what shaped the outcome.
Inspect the context, authority, policy, decisions, models, tools, human approvals, attempts, and external effects that became one business outcome.
Connect runtime evidence to application meaning
Execution context
Which Execution Instance, application version, proposal, context, and authority shaped the outcome?
Decision path
Which policy decision, routing record, model result, validation, and human approval participated?
Business effect
Which Business Operation Identity, attempts, reconciliation, and Effect Receipt explain the final state?
Keep one connected evidence spine across the application
Trace ID, proposal ID, policy decision ID, approval ID, Attempt ID, Business Operation Identity, and receipt ID solve different problems. Frozion preserves the relationships among them.
Declare how faithful reconstruction can be
Historical replay can identify what is preserved, resolved from versioned artifacts, reconstructed, substituted, simulated, or unavailable.
Keep replay isolated from production effects
Side-effecting nodes can be denied, stubbed, substituted, sandboxed, or simulated so investigation does not change business state.
Keep application causality precise
Behavior Intelligence should not overclaim access to private model reasoning. It explains the application-visible decision boundary.
Inspect one execution or compare many
Single execution
Why did this supplier change require human review, which policy applied, and what effect was confirmed?
Cross-execution analysis
Which route causes escalation, which obligation causes waits, and which capability produces abstentions?
Scoped projections
Analyze by tenant, application version, graph node, policy, capability, model, route, reviewer, and outcome.
Choose replay mode explicitly
Inspect-only
Read preserved evidence and realized execution path.
Historical reconstruction
Rebuild historical graph, state, context, decisions, waits, attempts, and effects.
Counterfactual simulation
Run a historical case against alternate policy, model, routing, threshold, or capability.
Controlled re-execution
Permit a current action only after present-day safety gates.
Start from one business outcome and reconstruct what shaped it.
Use an architecture review to identify the application identities, evidence, replay modes, and safety boundaries needed for one consequential AI process.