State model
A partitioned event log feeds time-windowed operational state. Per-entity sequence and watermark prevent late or duplicate events from appearing current.
Potential conduct issues span orders, trades, communications, market events and employee relationships, with many legitimate alternative explanations.
The system aligns temporal evidence, proposes competing hypotheses and produces an investigation-ready case without assigning intent.
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.
Typed commands, state changes and receipts cross an event spine without surrendering domain ownership.
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.
Events or conversation turns require a decision before the underlying situation changes, with strict latency and back-pressure constraints.
Surveillance signals cannot trigger disciplinary action or external reporting without independent investigation.
A deterministic outer workflow contains model-led work inside typed, observable calls. Dashed messages remain proposals until policy or a human grants authority.
A partitioned event log feeds time-windowed operational state. Per-entity sequence and watermark prevent late or duplicate events from appearing current.
Use streaming ingestion for signals, low-latency feature or state reads for the hot path, and asynchronous enrichment outside the decision budget.
Point-in-time correctness takes precedence over the newest unqualified value. Late events trigger correction or review instead of mutating the old decision invisibly.
These roles are deliberately vendor-neutral. Each can be independently owned, versioned and replaced.
Returns candidate entities and typed relationships with match features, contradictions, effective dates and non-merge evidence.
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.
Builds time-qualified projections from source events and reconciliations without becoming the legal system of record.
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.
Bias consequential journeys against false merge and retain unresolved candidates.
Automatically merge the highest-scoring candidate.
Cost acceptedMore cases require clarification, but one person's authority or risk cannot silently attach to another.
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.
Use event-fed projections for scale and direct readback for consequential effects.
Fan out to all systems of record for every interaction.
Cost acceptedRead models introduce lag and reconciliation work, but reduce source load and make cross-system views feasible.
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.
partition_rate = peak_events_per_second / active_partitionsdecision_budget = ingest + state_read + policy + score + action_commitbacklog_clear_time = queued_events / recovery_throughputSeparate 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
Time alignment, source completeness, relationship paths, hypothesis support, counter-evidence and investigator decisions.
Conduct investigators determine escalation and outcome.
Start with advisory intervention and measured shadow scoring. Increase automation only when peak-load, late-event and fallback tests preserve the control outcome.
First-line control, second-line risk, compliance, audit and legal owners keep their separate decision rights.