{"schema_version":1,"assessment_id":"eoa_inverse_innovation_exp03_opportunity320_20260801","source_experiment_id":"eoa_inverse_innovation_exp03_full320_20260801","cell_id":"computability_boundary_mapping__accounting_auditing","archetype_slug":"computability_boundary_mapping","domain_slug":"accounting_auditing","title":"Scope-Relative Guarantees for Audit Automation","opportunity_summary":"Evaluate an offline assurance architecture that admits only proved bounded or decidable audit-analysis inputs, routes other inputs to labelled UNKNOWN or escalation, and prevents timeouts or missing witnesses from becoming clearance. The opportunity is conditional because the packet does not establish that audit stakeholders actually demand unrestricted, exact, terminating analysis.","adopter_authorizer":"The audit methodology owner can authorize an offline pilot; engagement leadership retains authority over audit conclusions, while financial-reporting management retains responsibility for the statements.","scores":{"meaningful_impact":{"score":4,"rationale":"If unrestricted analyzer claims are being used, preventing nontermination or incomplete search from becoming false clearance could materially improve truthful assurance. The packet does not establish how often that situation occurs."},"stakeholder_pull":{"score":2,"rationale":"The candidate identifies relevant actors and a plausible harm but provides no evidence that audit teams currently demand the specified universal exact-termination guarantee. Its own problem falsifier allows that stakeholders may seek only bounded-instance analysis."},"incremental_advantage":{"score":4,"rationale":"Compared with a Boolean analyzer that handles timeouts informally, enforced admission, explicit UNKNOWN routing, guarantee labels, and versioned evidence provide a clear and testable improvement in claim discipline. Advantage over conventional human-reviewed risk-based practice remains uncertain."},"distinctiveness_plausibility":{"score":3,"rationale":"The composition of computability classification, enforced scope admission, labelled routing, and audit-specific authority controls is internally differentiated, but prior art is explicitly unsearched and world distinctiveness cannot be inferred closed-book."},"technical_implementability":{"score":3,"rationale":"Formalizing one property and enforcing one bounded fragment in an offline pilot appears feasible, but producing an independently checked impossibility result, preserving semantic fidelity, and enforcing admission across ordinary cases may be difficult."},"adoption_authority_feasibility":{"score":4,"rationale":"The packet identifies a pilot authorizer, preserves engagement and management authority, excludes audit-opinion changes, and supplies halt and rollback conditions. Broader organizational acceptance of UNKNOWN outputs is untested."},"evidence_readiness":{"score":3,"rationale":"The proposal supplies explicit problem and intervention falsifiers plus a bounded offline comparison using synthetic and previously adjudicated cases. It remains a hypothesis with no demonstrated test corpus, validated encoding, reviewed proof, or observed baseline error rate."},"safety_net_benefit":{"score":4,"rationale":"Explicit UNKNOWN states, guarantee labels, escalation, prohibited uses, and rollback directly reduce the risk that incomplete analysis becomes audit clearance. Benefits depend on labels surviving workflow pressure and UNKNOWN not being operationally coerced into clearance."},"scalability":{"score":2,"rationale":"Each property may require specialized encoding, fragment design, proof review, admission controls, and semantic validation. The restricted fragment may exclude ordinary material cases, while frequent UNKNOWN outputs could shift work into uncontrolled channels."}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"Design and execute one offline comparison for one formalized audit property using synthetic and a limited set of previously adjudicated cases, including adversarial admission tests and independent review of the formal argument.","confidence":"LOW","assumptions":["A small adjudicated case set is legally and operationally accessible.","The study uses existing analysis infrastructure rather than building a production platform.","Audit-domain and formal-methods specialists contribute substantial labor.","No audit opinions or live posting paths are affected."]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build an audit-grade prototype with enforced scope admission, bounded-fragment analyzer, UNKNOWN and escalation routing, guarantee labels, evidence versioning, access controls, validation, training, and limited workflow integration.","confidence":"LOW","assumptions":["Deployment remains limited to a small number of properties and engagements.","Existing identity, case-management, and evidence-retention systems can be integrated.","Independent methodology, legal, compliance, and proof review are required.","The prototype supplements rather than replaces human-reviewed procedures."]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"Launch across multiple engagement teams or properties with governed model and property maintenance, integration testing, monitoring, documentation, training, quality review, and formal change control.","confidence":"LOW","assumptions":["Material expansion beyond the initial bounded property is required.","Each added property needs domain-specific encoding and validation.","Audit-grade traceability, segregation of duties, and compliance review are mandatory.","No estimate is made for enterprise-wide or cross-firm standardization."]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Maintain approved fragments, proofs, guarantee labels, integrations, adjudicated test suites, access controls, monitoring, staff training, escalation review, and periodic independent validation for a limited operational program.","confidence":"LOW","assumptions":["The number of supported properties remains limited.","Specialist audit and formal-methods capacity is retained.","Software infrastructure is largely reused after launch.","Major regulatory redesign or enterprise-wide expansion is excluded."]}},"research_burden":"HIGH","earliest_credible_horizon":"3_TO_12_MONTHS","pipeline_gates":{"recognizable_externally_supportable_problem":{"status":"UNCERTAIN","reason":"The failure mode is precisely stated and falsifiable, but the sealed packet provides no evidence that real audit stakeholders demand unrestricted exact terminating analysis or misreport timeouts and finite tests as universal assurance."},"identifiable_adopter_or_authorizer":{"status":"YES","reason":"The audit methodology owner is explicitly authorized to approve the offline pilot, with engagement leadership and management retaining their respective decision authority."},"distinct_testable_incremental_claim":{"status":"YES","reason":"The candidate claims that enforced boundary mapping, labelled UNKNOWN routing, and escalation will reduce overstated clearances relative to the Boolean baseline without materially increasing missed actionable cases under an agreed bounded test."},"bounded_next_evidence_step":{"status":"YES","reason":"The authorized first step is confined to one formalized property, one bounded fragment, synthetic and previously adjudicated cases, adversarial in-scope and out-of-scope inputs, and a baseline comparison without issuing audit opinions."},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The pilot is offline; live posting, opinion changes, UNKNOWN-as-clearance, silent exclusion, and unchecked impossibility claims are prohibited. Explicit halt conditions restore existing human-reviewed procedures."},"implementation_cost_scope_and_range":{"status":"YES","reason":"The one-property offline pilot and the principal labor, data, review, compliance, integration, and evaluation components are sufficiently bounded to assign broad resource-equivalent ranges, although confidence is low for scaled deployment."}},"blocking_evidence":["Evidence that audit teams actually seek unrestricted exact terminating analysis or currently convert timeouts, finite tests, or missing witnesses into universal clearance.","A validated encoding showing that the selected program property corresponds closely enough to the audit concern and does not omit economically material behavior.","An independently reviewed constructive classification or impossibility argument for the selected decision class, including its exact quantifiers and scope.","A representative adjudicated test set containing in-scope, out-of-scope, and adversarial cases, with agreed definitions of overstated clearance and missed actionable cases.","Evidence that the restricted fragment covers enough ordinary material cases and that UNKNOWN outputs remain visible and trigger accountable escalation.","Prior-art research establishing whether the proposed composition adds a meaningful increment over existing formal-methods, automated-control-testing, or assurance-routing practices."],"next_evidence_step":"Run a preregistered offline study for one audit property: first verify through structured review of recent methodology artifacts and adjudicated cases that the unrestricted-assurance failure mode occurs; then compare the Boolean baseline with enforced scope admission and labelled UNKNOWN routing on a fixed synthetic and retrospective corpus. Stop or reject the opportunity if no universal claim is found, any in-scope wrong clearance occurs, admission can be bypassed, labels are lost, overstated clearances do not decrease, or missed actionable cases materially increase under the agreed threshold.","research_questions":["Do methodology documents, analyzer requirements, or adjudicated cases contain unrestricted exact-termination claims, and are timeouts or absent witnesses ever treated as clearance?","What precise audit property, encoding, execution model, and quantifiers define the first decision class?","Is that class constructively decidable, undecidable under an independently checked reduction, or merely computationally difficult within relevant bounds?","How accurately does the encoded property represent external facts, estimates, legal interpretations, and materiality judgments relevant to the audit conclusion?","What proportion of ordinary material cases is admitted by the bounded fragment, and what work shifts to UNKNOWN or uncontrolled channels?","Compared with the Boolean baseline, how many overstated clearances and missed actionable cases does the router produce on a fixed adjudicated test set?","Can users bypass admission, suppress guarantee labels, or coerce UNKNOWN into clearance under realistic workflow pressure?","What existing formal-methods audit tools, continuous-assurance protocols, and scoped control-testing practices already implement similar mechanisms?"],"recommendation":"VALIDATE_PROBLEM_FIRST","uncertainty_constraints":["Closed-book assessment provides no evidence of problem prevalence, stakeholder demand, market size, realized impact, or adoption behavior.","Prior art is unsearched, so distinctiveness and incremental advantage relative to existing practice remain unmeasured.","The strongest countercase is that real engagements ask only finite-period, bounded-instance questions, making complexity, evidence quality, and judgment more relevant than computability.","A proof about encoded program behavior cannot establish financial-statement truth unless semantic fidelity to external facts and judgments is separately demonstrated.","Cost bands are resource-equivalent planning ranges based only on the stated scope and exclude unknown enterprise-wide, regulatory, litigation, and cross-firm standardization requirements.","Evidence from one property or bounded fragment would not justify generalization to arbitrary audit automation."] ,"closed_book_prior_art_boundary":"No claim is made about novelty, prevalence, existing products, market size, or realized outcomes. Prior art is explicitly unsearched; external research must compare the proposed mechanism composition with relevant formal-methods audit tools, continuous-auditing assurance protocols, and scoped automated-control testing before any distinctiveness claim."}