Home · Writing · Deployment

Population-Scale Outcome Testing for Customer Interactions

An evidence architecture for moving from small manual samples to risk-based testing across calls, chats and complaints while keeping customer-outcome and conduct judgement with accountable reviewers.

TLDR

  1. An evidence architecture for moving from small manual samples to risk-based testing across calls, chats and complaints while keeping customer-outcome and conduct judgement with accountable reviewers.
  2. A financial-services firm may handle millions of calls, chats and complaints while reviewing only a small manual sample.
  3. The control design begins with a test point: what behaviour or outcome is being examined, which population is in scope, what evidence is required and who decides.
  4. A conversation rarely stands alone. The customer may have contacted the firm repeatedly. An agent may promise an action that completes later.
  5. The transcript is a derived artefact. Material findings should retain audio time offsets or exact message IDs so a reviewer can inspect the source.
Figure 1Calls, chats and case records to actions, themes and assurance evidenceCausal and control schematic
Calls, chats and case records to actions, themes and assurance evidence6 declared states connected by 5 authored relations. The figure supports the section Coverage is not the same as assurance. L0L1L2L3L4 01
Calls, chats and case records
02
Evidence preparation
03
Deterministic and model tests
04
Risk-based reviewer routing
05
Authorised outcome assessment
06
Actions, themes and assurance evidence
Reading. The authored topology makes 5 declared relations across 6 states inspectable. Read it as the control structure for “Coverage is not the same as assurance”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.
On this page

Coverage is not the same as assurance

A financial-services firm may handle millions of calls, chats and complaints while reviewing only a small manual sample. Speech-to-text, classifiers and language models can examine a much larger population. That wider coverage can expose patterns that random sampling misses. It can also create false confidence if transcription error, weak labels or a persuasive summary becomes an “outcome score.”

Customer outcome is not directly observable in a transcript. The record may show disclosure, explanation, empathy, delay, vulnerability cues and action taken. A reviewer still needs policy, product, customer and case context to judge whether the interaction delivered the required outcome.

Models may detect, classify, summarise and rank. Accountable business and control roles retain customer-outcome judgement, breach determination, remediation and regulatory reporting.

Define the outcome question before selecting models

The control design begins with a test point: what behaviour or outcome is being examined, which population is in scope, what evidence is required and who decides. “Quality” is too broad. A mortgage explanation, fraud warning, complaint response and investment-support conversation have different obligations and evidence.

Test point Observable evidence Additional context Decision owner
Required disclosure Phrase, timing, acknowledgement Applicable script or policy version Quality or compliance reviewer
Customer understanding Questions, corrections, confirmations Product complexity and customer needs Authorised reviewer
Vulnerability response Cue, offered adjustment, channel treatment Known needs and permitted handling Customer-outcome specialist
Complaint recognition Dissatisfaction, harm or requested remedy Complaint definition and prior contacts Complaints function
Advice boundary Recommendation-like language and action User role, product and regulated service Compliance or advice supervisor
Resolution Promised and completed action Case state and postcondition evidence Operations owner

The test contract records inclusion, exclusion, sampling or coverage mode, labels, tolerances, evidence and escalation. It also defines legitimate unknown outcomes. A missing recording, incomplete transcript or unavailable case state should not receive a passing score.

Figure 2Policy, control and outcome objective to execution planCausal and control schematic
Policy, control and outcome objective to execution plan6 declared states connected by 7 authored relations. The figure supports the section Define the outcome question before selecting models. L0L1L2L3 01
Policy, control and outcome objective
02
Versioned test contract
03
Population query
04
Label and evidence specification
05
Reviewer and escalation rules
06
Execution plan
Reading. The authored topology makes 7 declared relations across 6 states inspectable. Read it as the control structure for “Define the outcome question before selecting models”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The FCA Consumer Duty is outcomes focused and requires firms to assess whether they deliver good outcomes. It does not prescribe one interaction classifier. The FCA's July 2026 Mills Review anticipates AI-driven transformation in retail finance while emphasising accountability, consumer outcomes and risk-tiered automation. Those sources should shape control questions, not become generic prompt text.

Build a multi-record interaction case

A conversation rarely stands alone. The customer may have contacted the firm repeatedly. An agent may promise an action that completes later. A warm handover may split one journey across chat and voice. Outcome testing should therefore build an interaction case with stable identifiers and bounded time.

Figure 3Voice segments to evidence manifest and chronologyCausal and control schematic
Voice segments to evidence manifest and chronology7 declared states connected by 6 authored relations. The figure supports the section Build a multi-record interaction case. L0L1L2 01
Voice segments
02
Interaction case
03
Chat messages
04
Contact and customer context
05
Actions and postconditions
06
Applicable policy and script
07
Evidence manifest and chronology
Reading. The authored topology makes 6 declared relations across 7 states inspectable. Read it as the control structure for “Build a multi-record interaction case”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The case separates actor, customer, channel, contact reason, product, relevant vulnerability handling, actions, promises and completion. It also carries consent and privacy treatment. Recordings and transcripts are sensitive; test data should be minimised and access purpose-bound.

Evidence object Authority Limitation
Audio recording Captures the audible interaction May omit screen actions or off-call work
Transcript Searchable representation of audio Contains recognition and diarisation error
Chat log Exact recorded messages for that channel May not show upstream bot or system state
CRM note Staff member's recorded summary Observation, not a verbatim source
Policy or script Defines expected treatment for scope and date Does not prove delivery or understanding
Action receipt Shows system state changed Does not prove the change met customer need

The transcript is a derived artefact. Material findings should retain audio time offsets or exact message IDs so a reviewer can inspect the source.

Treat speech processing as its own evaluated system

Speech-to-text quality varies by accent, noise, code switching, channel compression, names and specialist vocabulary. Aggregate word error rate can hide failure on the phrases that matter, such as a monetary amount, negation or vulnerability cue.

Figure 4Audio and channel metadata to downstream outcome testsCausal and control schematic
Audio and channel metadata to downstream outcome tests7 declared states connected by 6 authored relations. The figure supports the section Treat speech processing as its own evaluated system. L0L1L2L3L4 01
Audio and channel metadata
02
Diarisation and segmentation
03
Speech recognition
04
PII-aware transcript representation
05
Domain phrase and material-field checks
06
Transcript with quality map
07
Downstream outcome tests
Reading. The authored topology makes 6 declared relations across 7 states inspectable. Read it as the control structure for “Treat speech processing as its own evaluated system”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

Evaluation should include material phrase accuracy, speaker attribution, amount and date accuracy, negation, overlap and omitted audio. Low-quality regions route to playback or exclusion. The downstream model sees the quality map and must not state that a phrase was absent when the relevant audio was unintelligible.

The architecture also needs tamper and completeness checks. A recording start or stop failure, unexpected silence or missing segment becomes an exception. It cannot be treated as a clean interaction.

No adverse customer-outcome finding may rely only on an unreviewed low-quality transcript segment. The system must provide the source audio or route the case as insufficient evidence.

Combine deterministic tests, classifiers and language models

Not every test point needs a generative model. Required phrases, sequence, duration and completed system actions may be checked deterministically. Domain classifiers can identify topics, complaint cues or likely vulnerability signals. A language model can assemble cited evidence and compare it with a test contract.

Figure 5Interaction case to reviewer queueCausal and control schematic
Interaction case to reviewer queue7 declared states connected by 8 authored relations. The figure supports the section Combine deterministic tests, classifiers and language models. L0L1L2L3L4 01
Interaction case
02
Deterministic sequence and action checks
03
Approved classifiers
04
Evidence-grounded language analysis
05
Finding candidates
06
Cross-check and conflict service
07
Reviewer queue
Reading. The authored topology makes 8 declared relations across 7 states inspectable. Read it as the control structure for “Combine deterministic tests, classifiers and language models”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.
Method Suitable work Important limit
Rules Required phrase, order, timing, action receipt Synonyms and contextual meaning
Speech or text classifier Known cue or topic detection at scale Label drift and segment dependence
Anomaly detection Unusual pattern or team shift Anomaly is not poor outcome
Language model Evidence synthesis and nuanced comparison Unsupported judgement and sensitivity to context
Human review Material contextual and conduct judgement Capacity, consistency and anchoring

The certainty gradient should determine routing. High-confidence, low-consequence passes can reduce manual review if validation supports it. Suspected failures, vulnerable-customer cases, severe outcomes and uncertain evidence receive human review. Some findings always require review regardless of score.

The model's confidence is not the customer's outcome. Calibration can help route work; it cannot confer authority.

Use risk-based sampling on top of broad screening

Population screening can rank or stratify interactions, but assurance also needs unbiased coverage. If reviewers see only model-selected cases, false negatives remain invisible and the model's blind spots define the control.

Figure 6Full eligible population to performance and assurance estimatesCausal and control schematic
Full eligible population to performance and assurance estimates8 declared states connected by 10 authored relations. The figure supports the section Use risk-based sampling on top of broad screening. L0L1L2L3 01
Full eligible population
02
Automated screening
03
High-risk targeted sample
04
Uncertain and novel sample
05
Random representative sample
06
Control and known-negative sample
07
Reviewer assessment
08
Performance and assurance estimates
Reading. The authored topology makes 10 declared relations across 8 states inspectable. Read it as the control structure for “Use risk-based sampling on top of broad screening”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The sample design should retain strata by product, channel, journey, customer group, team, language, outcome severity and model confidence. Random samples estimate population performance. Targeted samples improve defect discovery. Disagreement samples identify label or boundary weakness. The proportions and analysis should be explicit.

Sample stream Purpose What it estimates
Random Detect false negatives and population trend Unbiased outcome rate within design limits
High risk Find likely material cases quickly Case discovery, not prevalence
Uncertain Improve boundaries and reviewer guidance Ambiguity profile
Novelty Detect new language, products or journeys Emerging risk signals
Known controls Verify stable positive and negative behaviour Regression and pipeline health

Reviewers should not see the model's final label before making an independent judgement on calibration samples. This reduces anchoring and allows honest agreement estimates.

Make findings evidence-linked and contestable

A finding candidate states the test point, observed evidence, source offsets, model or rule version, counterevidence, uncertainty and proposed severity. The reviewer can accept, amend, reject or request evidence. Rejected candidates remain useful for monitoring false positives.

Figure 7Finding candidate to remediation and thematic analysisCausal and control schematic
Finding candidate to remediation and thematic analysis8 declared states connected by 7 authored relations. The figure supports the section Make findings evidence-linked and contestable. L0L1L2L3L4
Accept
Amend
Reject
Insufficient
01
Finding candidate
02
Openable evidence and counterevidence
03
Reviewer decision
04
Outcome or control finding
05
Corrected finding with reason
06
False-positive record
07
Evidence gap and follow-up
08
Remediation and thematic analysis
Reading. The authored topology makes 7 declared relations across 8 states inspectable. Read it as the control structure for “Make findings evidence-linked and contestable”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The reviewer interface should lead with the relevant segment and policy requirement, not a long summary. It should show what the system did not consider. For a broken promise, the evidence includes both the spoken commitment and the downstream action record. For an advice-boundary concern, it includes the precise language and the staff member's authorised role.

Outcome and breach states remain separate. A poor outcome may require remediation without establishing a regulatory breach. A control failure may not have harmed a customer in the sampled case. The data model should not collapse them.

Evaluate models, trajectories and review operations

The evaluation set needs hard negatives, indirect language, multiple speakers, interrupted journeys, policy changes, vulnerability cues, sarcasm, code switching, ASR errors and delayed postconditions. It should include cases where the right answer is insufficient evidence.

Figure 8Expert-labelled case set to release scorecardCausal and control schematic
Expert-labelled case set to release scorecard7 declared states connected by 10 authored relations. The figure supports the section Evaluate models, trajectories and review operations. L0L1L2 01
Expert-labelled case set
02
Speech evidence tests
03
Classifier and rule tests
04
Language-analysis tests
05
Trajectory and routing tests
06
Human-review tests
07
Release scorecard
Reading. The authored topology makes 10 declared relations across 7 states inspectable. Read it as the control structure for “Evaluate models, trajectories and review operations”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.
Layer Measure Material failure
Speech Material phrase and speaker accuracy Negation or amount reversed
Detection Recall and precision by test point and segment Vulnerability cue missed in one channel
Evidence Citation accuracy and counterevidence coverage Finding points to wrong part of call
Routing Correct severity, abstention and reviewer path Severe case auto-passed
Review Inter-reviewer agreement and override quality Anchored acceptance without evidence inspection
Population Weighted estimates with sample design Targeted sample reported as population prevalence

The April 2026 US interagency model risk guidance explicitly excludes generative and agentic AI from its model scope while stating that other governance should determine appropriate controls. A population-testing system therefore needs an integrated governance pack across speech, classifiers, language models, rules, workflow and human decision, not a single model card.

Production monitoring tracks transcription quality, missing recordings, finding rate by segment, reviewer disagreement, false-negative samples, policy version, routing drift, queue ageing and remediation recurrence. A change in call mix can invalidate historical thresholds.

Connect findings to improvement without creating surveillance

The purpose of outcome testing is to improve customer treatment and control effectiveness. Findings should aggregate into themes with sufficient evidence and uncertainty. Individual staff use must follow clear policy, notice, access and fairness controls. A system designed for customer assurance should not quietly become an opaque performance-ranking tool.

Figure 9Reviewed findings to outcome monitoringCausal and control schematic
Reviewed findings to outcome monitoring6 declared states connected by 8 authored relations. The figure supports the section Connect findings to improvement without creating surveillance. L0L1L2L3 01
Reviewed findings
02
Evidence-backed themes
03
Policy and process improvement
04
Targeted coaching
05
System or journey repair
06
Outcome monitoring
Reading. The authored topology makes 8 declared relations across 6 states inspectable. Read it as the control structure for “Connect findings to improvement without creating surveillance”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

Access to recordings, transcripts and findings should be purpose-bound. Retention follows the relevant record policy. Evaluation datasets require minimisation, lawful use and separation from general model training. Reviewer corrections do not automatically become training labels without quality and privacy controls.

The NIST AI RMF provides a useful govern, map, measure and manage structure. It should be instantiated in the operating process: named owners, data and system maps, disaggregated measures, incident response and documented risk treatment.

A measured path to production

Start with one outcome question and one channel where policy, evidence and reviewer ownership are clear. Run beside the existing sample process. Reconcile every disagreement. Prove that source links, postconditions and unknown outcomes work before expanding population coverage.

Then add test points and channels in controlled releases. Use independent random sampling to estimate blind spots. Adjust reviewer capacity before widening automated screening. Release thresholds should be set by consequence and evidence, not by a universal model score.

Figure 10One test point and channel to thematic improvement and assuranceCausal and control schematic
One test point and channel to thematic improvement and assurance6 declared states connected by 5 authored relations. The figure supports the section A measured path to production. L0L1L2L3L4 01
One test point and channel
02
Shadow screening
03
Independent reviewer reconciliation
04
Risk-based production routing
05
Broader population and channels
06
Thematic improvement and assurance
Reading. The authored topology makes 5 declared relations across 6 states inspectable. Read it as the control structure for “A measured path to production”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The value case measures reviewed coverage, time to material finding, false-negative discovery, evidence-preparation effort, repeated harm, remediation closure and customer-outcome improvement. It should not claim success from transcripts processed or alerts generated.

Work one outcome case across channels

Consider whether a customer in financial difficulty received appropriate support across a call, two secure messages, a payment arrangement and a later complaint. No individual record proves the outcome. The call contains uncertainty about income. The message offers several options. The payment system records an arrangement. The complaint later says the customer did not understand a consequence.

The case builder first resolves the interaction boundary using customer, product, journey and time. It records source availability before interpretation. If one recording is missing, the case is incomplete rather than clean. Speech processing produces time-aligned words, speakers and confidence. Policy retrieval selects the relevant test points by product, customer state and interaction time.

Figure 11Interaction and transaction sources to remediation and thematic learningCausal and control schematic
Interaction and transaction sources to remediation and thematic learning8 declared states connected by 8 authored relations. The figure supports the section Work one outcome case across channels. L0L1L2L3L4 01
Interaction and transaction sources
02
Time-bounded customer case
03
Applicable outcome test points
04
Evidence propositions and gaps
05
Hybrid screening
06
Prioritised human review
07
Finding with counterevidence
08
Remediation and thematic learning
Reading. The authored topology makes 8 declared relations across 8 states inspectable. Read it as the control structure for “Work one outcome case across channels”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

Deterministic checks identify whether required disclosures and actions occurred. A classifier locates likely vulnerability and comprehension cues. A language model assembles cited evidence for and against each test point. It may state that the customer expressed uncertainty and that the staff member offered options. It cannot infer comprehension merely from the absence of an objection.

The reviewer sees the original audio interval, transcript, messages, payment record, policy point and contradictory evidence. The finding separates process adherence, likely customer understanding, downstream outcome and remaining uncertainty. An interaction outcome is a reasoned case conclusion, not a sentiment score attached to a call.

Test fairness across segments, channels and evidence quality

Aggregate precision can hide harmful differences. Evaluation should be disaggregated by channel, language variety, accent, recording quality, product, vulnerability indicator, interaction length and journey type where lawful and statistically defensible. The purpose is to find performance and treatment gaps, not to create unsupported personal classifications.

Slice Question Guardrail
Audio quality Does material-cue recall fall on poor recordings? Report uncertainty; do not infer absence
Language or accent Are transcription and speaker errors uneven? Expert review and targeted improvement
Channel Does message-only evidence receive different treatment? Channel-specific test design
Product Are policy points and thresholds correctly selected? Versioned applicability rules
Vulnerability context Are support cues found without over-labelling? Evidence-linked proposition and human review
Interaction length Does context truncation omit late disclosures? Coverage checks and chunk lineage
Reviewer group Does disposition vary after controlling for case mix? Calibration and independent adjudication
Figure 12Expert-reviewed evaluation cases to release decisionCausal and control schematic
Expert-reviewed evaluation cases to release decision7 declared states connected by 10 authored relations. The figure supports the section Test fairness across segments, channels and evidence quality. L0L1L2 01
Expert-reviewed evaluation cases
02
Overall performance
03
Channel slices
04
Speech-quality slices
05
Journey and product slices
06
Customer-context slices
07
Release decision
Reading. The authored topology makes 10 declared relations across 7 states inspectable. Read it as the control structure for “Test fairness across segments, channels and evidence quality”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

Small groups require confidence intervals and suppression rules. A raw percentage from a few cases should not become a fairness conclusion. Where protected attributes are unavailable or inappropriate, the programme can still test operational proxies such as channel, transcription quality and service route while documenting the limitation.

Thresholds may differ by test point because consequence and base rate differ. They should not vary in a way that reduces protection for a customer group. Any threshold change needs replay on stable cases, impact analysis and owner approval. Fairness assurance belongs in the release gate and production monitoring, not in an annual appendix.

Verify postconditions beyond the interaction

Many outcome questions depend on what happened after the conversation. A promised refund, forbearance action, complaint referral or accessibility adjustment must be verified in the relevant system. The language model cannot establish completion from a staff member's statement alone.

Figure 13Promise or required action detected to evidence gap and follow-upCausal and control schematic
Promise or required action detected to evidence gap and follow-up7 declared states connected by 6 authored relations. The figure supports the section Verify postconditions beyond the interaction. L0L1L2L3L4
Yes
No
Unknown
01
Promise or required action detected
02
Expected postcondition
03
Authoritative system query
04
Observed within allowed time?
05
Verified outcome evidence
06
Potential failure
07
Evidence gap and follow-up
Reading. The authored topology makes 6 declared relations across 7 states inspectable. Read it as the control structure for “Verify postconditions beyond the interaction”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The postcondition contract names system, identifier, state, time limit and safe retry. It distinguishes proposed, submitted, completed, reversed and unknown. An action recorded in notes is not equivalent to an executed servicing state. A payment arrangement that was created and immediately cancelled may require a different outcome conclusion from one that remained active.

Cross-system matching needs its own quality controls. Customer and case identifiers, timestamps and product scope should reconcile deterministically where possible. Probabilistic matching creates a review item rather than silently attaching another customer's event. Evidence manifests preserve the queried interval and source version.

This approach also improves remediation. If outcome testing finds that a promised action was not completed, the workflow can create a proposed correction under human approval and verify the result. It does not let the testing agent edit a customer account directly. Detection authority and remediation authority remain separate.

Calibrate reviewers and measure decision quality

Human review is not a perfect reference by default. Reviewers can interpret policy differently, overlook counterevidence or anchor on the model's suggested finding. Calibration should therefore use blind independent review, structured disagreement and adjudication by authorised domain experts.

Figure 14Calibration case to guidance, interface or model correctionCausal and control schematic
Calibration case to guidance, interface or model correction6 declared states connected by 6 authored relations. The figure supports the section Calibrate reviewers and measure decision quality. L0L1L2L3L4 01
Calibration case
02
Reviewer A without model suggestion
03
Reviewer B with evidence interface
04
Disagreement analysis
05
Policy-owner adjudication
06
Guidance, interface or model correction
Reading. The authored topology makes 6 declared relations across 6 states inspectable. Read it as the control structure for “Calibrate reviewers and measure decision quality”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

Review forms separate evidence sufficiency, test-point disposition, severity and required action. Free text explains reasoning but does not replace typed fields. The interface shows supporting and contrary spans before any suggested conclusion. Reviewers can ask for missing evidence and record why a source is unreliable.

Measures include agreement by test point, direction of overrides, evidence-opening behaviour, time to decision and recurrence after feedback. A high agreement rate after exposing a strong model recommendation may indicate anchoring. Periodic blind cases help detect it. Quality review samples both passes and findings because false negatives are the central risk in broad screening.

Reviewer feedback routes to the responsible component. Transcript correction goes to speech operations. Wrong policy selection goes to applicability. Unsupported drafting goes to the evidence composer. Disagreement about the expected customer outcome goes to the policy owner. A single override count cannot tell the programme what to improve.

Run a closed remediation and learning loop

A confirmed finding becomes a governed case with owner, severity, customer impact, required action and due date. Similar findings may share a systemic cause, but item-level lineage remains. The platform can cluster themes and suggest a causal hypothesis; accountable owners accept the remediation plan.

Stage Evidence required Closure condition
Finding accepted Reviewed interaction and test-point rationale Named owner and impact assessment
Immediate protection Required customer or service action Verified postcondition
Root cause Process, policy, training, system or data evidence Challenge by control owner
Remediation Defined change and affected population Implementation receipt
Lookback Search logic, population and sample design Reviewed affected cases
Effectiveness Independent outcome measure after change Sustained improvement within range
Figure 15Confirmed finding to evidence-backed closureCausal and control schematic
Confirmed finding to evidence-backed closure7 declared states connected by 7 authored relations. The figure supports the section Run a closed remediation and learning loop. L0L1L2L3L4
Insufficient
Sustained
01
Confirmed finding
02
Immediate customer protection
03
Root-cause analysis
04
Controlled change
05
Affected-population lookback
06
Effectiveness testing
07
Evidence-backed closure
Reading. The authored topology makes 7 declared relations across 7 states inspectable. Read it as the control structure for “Run a closed remediation and learning loop”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The lookback query is versioned and reviewed. It should not assume that the detection model defines the full affected population. Independent sampling estimates cases the query missed. Closure evidence remains separate from the team that implemented the fix where assurance requires independence.

Architecture review should ask whether missing records can be mistaken for clean cases, whether speech uncertainty is carried into findings, whether postconditions are checked in authoritative systems, whether segment performance is visible, whether reviewers are protected from anchoring. Whether remediation closure proves an improved customer outcome.

Anchor the service in conduct and data governance

The operating design should start from the outcome, policy and lawful purpose that justify examining interactions. The FCA Consumer Duty provides an outcome-centred supervisory frame for UK retail financial services. The FCA's guidance on vulnerable customers is relevant when designing evidence, support and review. These sources do not prescribe one model or monitoring architecture; they shape the questions accountable firms must answer.

Data governance should define which interactions are in scope, who can access them, how long derived artefacts remain, and whether outputs may be used for coaching, control testing, complaint handling or employment decisions. These purposes are not interchangeable. A model trained or evaluated for one purpose should not be repurposed through a dashboard setting.

Figure 16Named assurance purpose to explicitly excluded secondary usesCausal and control schematic
Named assurance purpose to explicitly excluded secondary uses6 declared states connected by 5 authored relations. The figure supports the section Anchor the service in conduct and data governance. L0L1L2L3L4 01
Named assurance purpose
02
Authorised population and sources
03
Analysis and reviewer access
04
Finding and remediation purpose
05
Retention and deletion policy
06
Explicitly excluded secondary uses
Reading. The authored topology makes 5 declared relations across 6 states inspectable. Read it as the control structure for “Anchor the service in conduct and data governance”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The ICO guidance on AI and data protection provides a useful basis for purpose, transparency, data minimisation and individual rights. Governance must cover recordings, transcripts, embeddings, case state, prompts, model traces, reviewer notes and evaluation samples. Removing the source audio while retaining a searchable derived representation is not deletion.

Role separation limits internal surveillance risk. Conduct assurance reviewers see what they need for the case. Line managers receive approved thematic or coaching outputs under policy. Model developers use curated and minimised evaluation data. Operations staff diagnose through metadata unless deeper access is authorised. Exports and downstream joins are logged and restricted.

Design production economics around reviewed outcomes

Population-scale processing can be expensive. The service should route work according to evidence need and consequence. Deterministic metadata checks run broadly. Speech and classifiers screen the eligible population. More expensive language-model analysis is invoked for uncertain or material cases. Human review remains focused but includes independent random samples.

Figure 17Eligible interaction population to human reviewCausal and control schematic
Eligible interaction population to human review7 declared states connected by 5 authored relations. The figure supports the section Design production economics around reviewed outcomes. L0L1L2L3 01
Eligible interaction population
02
Deterministic coverage checks
03
Speech and specialist classifiers
04
Risk and uncertainty routing
05
Monitored pass plus random sample
06
Evidence-linked language analysis
07
Human review
Reading. The authored topology makes 5 declared relations across 7 states inspectable. Read it as the control structure for “Design production economics around reviewed outcomes”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

Cost per reviewed and accepted outcome is more useful than cost per transcript. The model should record transcription, retrieval, analysis, evidence storage and review effort by test point and channel. Quality and cost are examined together. A cheap route that loses material cues creates false economy; an expensive route applied to every interaction can make the service unsustainable.

Economic measure What it reveals
Cost per eligible interaction Base population burden
Cost per material reviewed case Routing efficiency
Reviewer minutes by test point Evidence and interface quality
False-negative discovery cost Price of blind spots
Remediation value and recurrence Whether findings improve outcomes
Storage and replay cost Sustainability of evidence retention

Capacity tests should use peak call volumes, long interactions, poor recordings and correlated events such as product incidents. Queue limits and degradation policies are decided in advance. The service may defer low-consequence enrichment, but it should not silently reduce population coverage or evidence quality.

The business case should reward better detection, faster protection and sustained improvement, not the amount of content processed. That keeps investment connected to customer and control outcomes.

Make release decisions test-point specific

A service should not receive one broad approval for “conduct AI.” Each outcome test point has its own policy basis, source needs, error consequences, threshold, review route and remediation. A model may be reliable for disclosure presence and weak for comprehension. The first capability can proceed while the second remains under supervised study.

Figure 18Outcome test point to narrow, repair or withdrawCausal and control schematic
Outcome test point to narrow, repair or withdraw8 declared states connected by 7 authored relations. The figure supports the section Make release decisions test-point specific. L0L1L2L3L4
Stable
Defect
01
Outcome test point
02
Evidence and policy contract
03
Offline and shadow validation
04
Limited channel canary
05
Production monitoring
06
Authority review
07
Wider bounded coverage
08
Narrow, repair or withdraw
Reading. The authored topology makes 7 declared relations across 8 states inspectable. Read it as the control structure for “Make release decisions test-point specific”, not as measured performance. Schematic derived from the paper's authored topology; no measured quantities.

The release pack states eligible population, exclusions, source coverage, segment results, false-negative sampling, reviewer capacity, access policy and rollback triggers. It names what the system may do: screen, assemble evidence, recommend review priority or create a proposed finding. It also names what remains prohibited, such as automatic adverse action, staff ranking or customer-account change.

Change impact is equally specific. A speech model update may affect amount and negation recognition across several test points. A policy version may affect only one product. A reviewer-interface change may alter anchoring. The platform maps components to test points and replays the affected evaluation sets before release.

Rollback preserves case and evidence records. New screening can pause while accepted findings and remediation continue. Previously reviewed cases are not silently rescored. If a defect could have hidden material outcomes, the owner defines a lookback population and independent review.

Granular release authority makes the system easier to expand because every capability has a visible basis and reversal path. It also prevents a good result in one test from being used to justify unrelated automation.

The operating committee should review quality, customer impact, privacy, reviewer capacity, source coverage and cost in one view. Separate dashboards can make a technically improving model appear successful while queues age or remediation repeats. Decisions to widen a channel or test point should name the evidence and residual limits.

Front-line feedback is useful when it points to policy ambiguity, missing journey context or an impractical remediation route. It should not permit a team to suppress inconvenient findings. Feedback objects retain type, evidence and disposition. Material challenges go to an independent owner where the operating model requires it.

Customer corrections and complaints can reveal missed outcomes, but they are not a complete ground truth because many harms are never reported. The evaluation programme keeps independent sampling and targeted expert review. It also compares themes before and after policy, training or system changes so that closure reflects sustained behaviour.

This creates a disciplined path from broad visibility to specific action. The service remains valuable when it finds uncertainty, missing records and operating weakness, not only when it confirms compliance.

Population-scale outcome testing is credible when it expands visibility without hiding uncertainty or human responsibility. The platform's best contribution is not a single score. It is a reliable chain from interaction evidence to a reasoned, contestable and owned conclusion.