State model
A decision snapshot binds period, entity hierarchy, measures, formula versions, assumptions and source cut-offs. Scenarios never overwrite actuals.
A lower handling cost or shorter conversation does not prove that an AI route improved customer, control or operational outcomes.
The measurement design links route choice to service quality, rework, loss, complaints, control exceptions, human effort and delayed outcomes using comparable cohorts and explicit attribution limits.
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.
A shared authority core coordinates independently owned capability cells, domain systems and operating evidence.
Permitted workThe assurance layer observes and evaluates; it may trigger a stop or review but cannot redefine business policy from telemetry alone.
Consistency ruleCarry one correlation chain across synchronous and asynchronous hops and separate event occurrence from observation and processing time.
Hard boundaryThe model is not a system of record, identity provider, policy authority or proof that an external effect occurred.
A business operator needs a reconciled view, explanation or scenario from several data products while retaining ownership of the business decision.
Metrics are segmented by risk and journey complexity; the system cannot claim causality from simple before-and-after movement.
A deterministic outer workflow contains model-led work inside typed, observable calls. Dashed messages remain proposals until policy or a human grants authority.
A decision snapshot binds period, entity hierarchy, measures, formula versions, assumptions and source cut-offs. Scenarios never overwrite actuals.
Read from curated analytical products rather than operational tables. Use drill-through links for source detail and typed write-back only for approved decisions.
Definitions and units are resolved before aggregation. Missing data stays missing; mixed grains and double-counted entities fail reconciliation.
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.
Runs component, route, trajectory, failure, harm and outcome tests against the versioned system manifest.
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.
Carry one correlation chain across synchronous and asynchronous hops and separate event occurrence from observation and processing time.
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.
Test prompts, models, tools, knowledge, policies, state transitions and human paths together.
Use a static answer-quality benchmark as the release gate.
Cost acceptedSystem evaluation takes longer and needs synthetic environments, but detects authority and recovery failures that answer scoring misses.
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.
query_load = active_users x decisions_per_hour x source_scansscenario_work = entities x periods x assumptions x alternativesreview_value = avoided_rework + decision_gain - compute - human_effortCollect the minimum diagnostic metadata, tokenise subjects and keep raw prompts or case evidence behind stricter access and retention.
A design is production-ready only when teams can prove what happened, recover it and change it safely.
trace and outcome completeness
control-bypass and false-negative review
cost and human-effort attribution
kill-switch and recovery exercise
Route assignment, cohort definition, customer and control outcomes, downstream work, cost model, confidence limits and owner interpretation.
Finance, product, operations and risk owners agree the attribution method and decide investment.
Begin with explanation and transparent calculations. Add recommendation only where decision rights, counterfactual evaluation and outcome measurement are established.
Service, control, finance, model, data and operations owners interpret evidence and decide intervention.