State model
Batch and partition manifests remain immutable; reruns create new attempts. Published views point to the exact successful attempt and source cut-offs.
Risk and control self-assessment depends on fragmented incidents, metrics, control tests, changes and expert judgement across a defined period.
An evidence layer assembles the assessment population, identifies contradictions and drafts a rating rationale that keeps observation separate from judgement.
Organisations, systems and operating conditions are intentionally anonymised and recomposed. The design demonstrates engineering and banking-domain reasoning; it does not represent a named client estate, vendor product or measured production result.
Five operating planes separate interaction, identity, decision control, authoritative state and operating evidence.
Permitted workThe system assembles evidence and tests obligations. Accountable control owners decide ratings, exceptions, findings and closure.
Consistency ruleLink each conclusion to effective obligation, control, population, test and owner; corrections append rather than erase prior evidence.
Hard boundaryThe model is not a system of record, identity provider, policy authority or proof that an external effect occurred.
Several sources must agree on balances, positions, costs or outcomes before a reporting or management decision can rely on them.
The system cannot assign the final risk or control rating.
A deterministic outer workflow contains model-led work inside typed, observable calls. Dashed messages remain proposals until policy or a human grants authority.
Batch and partition manifests remain immutable; reruns create new attempts. Published views point to the exact successful attempt and source cut-offs.
Use file or event arrival to trigger orchestration. Separate raw landing, canonical transformation, control calculation and consumer serving.
A consumer sees one qualified batch version. Mixed source cut-offs or partial partitions cannot be presented as a complete period.
These roles are deliberately vendor-neutral. Each can be independently owned, versioned and replaced.
Accepts a typed evidence request and returns the minimum permitted facts with source, event time, observation time, validity and exclusions.
Appends request, versions, policy result, model proposal, approval, action receipt, readback, correction and custody events under one correlation key.
Chooses an approved model route by task, risk, evidence quality, latency budget and cost ceiling; enforces structured outputs.
Presents claims beside evidence, alternatives, uncertainty, missing information, permitted actions and current custody.
Evaluates identity, purpose, capability, amount, risk tier and policy version; returns allow, deny, step-up or human-review with reasons.
Durable records carry provenance, authority, effect and custody without turning a transcript into an uncontrolled memory store.
Link each conclusion to effective obligation, control, population, test and owner; corrections append rather than erase prior evidence.
The selected design is not universally superior. It is the safer fit for this boundary and failure cost.
Prefetch stable, purpose-safe facts; acquire volatile facts on demand against a time-qualified snapshot.
Load a broad customer profile at session start.
Cost acceptedThe selected design adds source calls and latency, but reduces stale data, excess exposure and accidental reuse.
Use append-only events plus a rebuildable current-state projection.
Overwrite the case row with its latest status.
Cost acceptedReplay and storage are more complex, but point-in-time reconstruction and correction lineage remain possible.
Reserve larger models for residual reasoning after deterministic and smaller-model gates.
Send every request to the most capable available model.
Cost acceptedRouting adds evaluation work and operational complexity, but controls cost, latency and unnecessary data exposure.
Keep people at irreversible, ambiguous and policy-exception points; sample lower-risk automated outcomes independently.
Require the same manual approval at every step.
Cost acceptedRisk-tiering reduces review load but needs calibrated thresholds, sampling and immediate withdrawal of authority when drift appears.
Compile stable decision logic and retain retrieval for explanation and residual ambiguity.
Ask a model to interpret the source document for every request.
Cost acceptedRule compilation needs controlled change, but creates repeatable decisions, regression tests and clear exceptions.
Retries are bounded by knowledge of business effect; unknown outcome remains visible, owned and independently reconciled.
Actual thresholds belong to accountable service owners. The design exposes the equations and observables that those owners must baseline.
required_throughput = input_rows / available_batch_windowreconciliation_work = partitions x independent_control_setsrerun_headroom = remaining_window x spare_processing_rateSeparate privileged, investigation and employee data from general model context; preserve legal-hold and access evidence.
A design is production-ready only when teams can prove what happened, recover it and change it safely.
obligation-to-control coverage
sampling and population integrity
evidence freshness and independence
finding closure and residual-risk approval
Population definition, source completeness, control results, incident links, rationale citations, challenge and signed rating.
Risk and control owners assign ratings and accept residual risk.
Move high-value changes to event-driven refresh only after the batch control set can reconcile the new path against the established baseline.
First-line control, second-line risk, compliance, audit and legal owners keep their separate decision rights.