MaxArc Global Health Impact Platform

Cross-Module AI Verification & Anomaly Intelligence

Cross-module signals · Evidence review · Authorized search · Human-decided

AI unifies the signals. Authorized humans make every decision.

A single cross-module intelligence layer over fund accountability, stockpile logistics, laboratory operations, patient continuity, restricted locator, asset management, and program performance. AI reconciles records, surfaces anomalies and contradictions, prioritizes review, and summarizes evidence — but AI confidence is not proof and an anomaly is never a finding until an authorized human reviews and decides. Patient-related signals stay privacy restricted; this is not a public search engine.

You are viewing the Cross-Module AI Verification & Anomaly Intelligence Dashboard
Evidence & decision posture: Every intelligence case references source records and evidence. AI confidence is not proof and anomaly signals are not findings until an authorized human reviews and decides. Critical decisions require authorized human approval; sensitive cases may require a second review. Requestor, reviewer, approver, and auditor roles remain separated where required. False-positive decisions require an explanation; suppressed or dismissed cases require a documented reason and audit history. Source records are never silently modified, contradictory evidence remains visible, and incomplete evidence prevents final certification. Patient-related signals remain privacy restricted. Every search, review, decision, escalation, override, and closure is auditable. These are operational AI-assisted signals, not confirmed clinical or compliance findings.
Open cases 15

Awaiting resolution

Critical & high 12

Highest priority

Awaiting review 15

Pending human decision

Overdue reviews 15

Past review due date

Second-review queue 6

Sensitive cases

SoD exceptions 1

Separation-of-duties flags

Evidence incomplete 10

Blocks certification

Cross-module contradictions 7

Conflicting sources

Unresolved reconciliation 15

Open / in progress

Suspected diversion/misuse 4

Under authorized review

Privacy-restricted signals 5

Patient-related

False-positive rate 50%

Of decided cases

Confirmed findings 1

Human-decided

Dismissed findings 1

Human-decided

Audit completeness 100%

Source + evidence referenced

Intelligence cases

Every case references source records and evidence. Confidence and evidence completeness are shown but do not constitute proof. Cases remain "pending human review" until an authorized reviewer decides.

CaseModuleSeverityAI signalsSource / evidenceStatus
AIC-0001
suspected grant misuse
fund-accountability critical conf 0.72
evid 65%
FA-DISB-2201, FA-VENDOR-118
evidence: EVD-FA-2201
pending human review 2nd review escalated
AIC-0002
procurement irregularity
fund-accountability high conf 0.6
evid 50%
FA-PROC-3390
evidence: EVD-FA-3390
pending human review
AIC-0003
commodity stock discrepancy
stockpile-logistics high conf 0.68
evid 70%
SL-BATCH-7781, SL-MOVE-9902
evidence: EVD-SL-7781
pending human review
AIC-0004
possible diversion
stockpile-logistics critical conf 0.74
evid 55%
SL-MOVE-9930
evidence: EVD-SL-9930
pending human review 2nd review escalated
AIC-0005
unexplained warehouse movement
stockpile-logistics medium conf 0.58
evid 60%
SL-WH-4410, SL-MOVE-9951
evidence: EVD-SL-4410
pending human review
AIC-0006
unusual laboratory result pattern
lab-operations privacy medium conf 0.61
evid 75%
LAB-RES-5521
evidence: EVD-LAB-5521
pending human review
AIC-0007
delayed doctor routing
lab-operations privacy high conf 0.66
evid 80%
LAB-ORD-6600
evidence: EVD-LAB-6600
pending human review
AIC-0008
unresolved lab result follow up
patient-continuity privacy high conf 0.7
evid 60%
PC-CASE-7010
evidence: EVD-PC-7010
pending human review
AIC-0009
treatment interruption risk
patient-continuity privacy critical conf 0.77
evid 70%
PC-CASE-7025
evidence: EVD-PC-7025
pending human review 2nd review escalated
AIC-0010
suspicious restricted locator access
restricted-patient-locator privacy critical conf 0.8
evid 85%
RPL-AUDIT-8100
evidence: EVD-RPL-8100
pending human review 2nd review escalated
AIC-0011
duplicate or ghost asset
asset-management high conf 0.69
evid 50%
AST-0007, AST-0001
evidence: EVD-AST-0007
pending human review
AIC-0012
unexplained asset movement
asset-management medium conf 0.62
evid 45%
AST-0004, TRF-3002
evidence: EVD-AST-0004
pending human review
AIC-0013
fabricated or duplicate program reporting
program-performance critical conf 0.73
evid 30%
RES-TB-02-Q2
evidence: EVD-PP-TB-02
pending human review 2nd review escalated
AIC-0014
denominator manipulation
program-performance high conf 0.64
evid 55%
RES-MAL-02-Q2
evidence: EVD-PP-MAL-02
pending human review
AIC-0015
cross module contradiction
cross-module critical conf 0.71
evid 50%
RES-MAL-02-Q2, SL-BATCH-7781, FA-DISB-2201
evidence: EVD-XM-0015
pending human review 2nd review escalated
AIC-0016
unexplained warehouse movement
stockpile-logistics low conf 0.41
evid 90%
SL-MOVE-9800
evidence: EVD-SL-9800
decided
AIC-0017
suspected grant misuse
fund-accountability high conf 0.7
evid 95%
FA-DISB-2150
evidence: EVD-FA-2150
decided 2nd review

Cross-module contradictions (evidence stays visible)

  • AIC-0001 (fund-accountability) — corroborating: procurement-record, bank-reconciliation; conflicting: vendor-attestation. Next: authorized reviewer reconciles disbursement against procurement and bank records before any decision
  • AIC-0003 (stockpile-logistics) — corroborating: stock-count, dispatch-record; conflicting: facility-receipt. Next: reconcile dispatched vs received quantities for batch
  • AIC-0006 (lab-operations) — corroborating: qc-log; conflicting: repeat-test. Next: lab quality officer reviews QC and repeat-test discordance; not a diagnosis
  • AIC-0011 (asset-management) — corroborating: registry-duplicate; conflicting: verification-record. Next: physical verification to resolve suspected duplicate/ghost asset
  • AIC-0013 (program-performance) — corroborating: improbable-increase; conflicting: disputed-verification. Next: obtain source documents before any reporting; do not certify
  • AIC-0015 (cross-module) — corroborating: program-result, stock-record, disbursement; conflicting: facility-receipt, vendor-attestation. Next: reconcile testing-coverage shortfall against RDT stockout and disbursement; authorized review only
  • AIC-0016 (stockpile-logistics) — corroborating: warehouse-log; conflicting: authorized-schedule. Next: no action; movement matched an authorized schedule

Reconciliation workflow

  • REC-0001 — dispatched 500 vs received 460 units in-progress
    modules: stockpile-logistics · reviewer: logistics-controller
  • REC-0002 — testing coverage below target while RDT stockout and disbursement recorded in-progress
    modules: program-performance, stockpile-logistics, fund-accountability · reviewer: oversight-reviewer
  • REC-0003 — two registry entries for one physical cold-chain unit in-progress
    modules: asset-management · reviewer: asset-controller

Human decisions (AI recommendation shown alongside)

Decisions are made by authorized roles with separation of duties and second review for sensitive cases. False-positive and dismissal decisions require a documented explanation and audit history.

  • AIC-0016dismissed by logistics-controller. AI recommended: low-severity; likely benign, verify schedule. Rationale: matched authorized after-hours schedule
  • AIC-0017confirmed-finding by grant-compliance-officer (2nd: oversight-reviewer). AI recommended: high-confidence anomaly; recommend authorized investigation. Rationale: bank, procurement, and audit-trail evidence corroborated misuse

Authorized operational search

Authorized operational search demonstration over synthetic data. Respects module and privacy restrictions and returns minimum-necessary information. Not a public search engine; never exposes patient identity or precise location.

TypeModuleReferenceDisclosure
grant fund-accountability Malaria grant (synthetic)
GRANT-2024-MAL-01
operational
procurement fund-accountability RDT procurement (synthetic)
FA-PROC-3390
operational
commodity stockpile-logistics RDT batch (synthetic)
SL-BATCH-7781
operational
warehouse stockpile-logistics Central warehouse (synthetic)
SL-WH-4410
operational
facility stockpile-logistics Riverside Health Center (synthetic)
FAC-RIVER
operational
lab-order lab-operations Lab order (synthetic)
LAB-ORD-6600
restricted
minimum-necessary; no patient identity
lab-result lab-operations Lab result (synthetic)
LAB-RES-5521
restricted
minimum-necessary; no patient identity
continuity-case patient-continuity Continuity case (masked)
PC-CASE-7010
restricted
masked reference only; no identity or precise location
restricted-locator-audit restricted-patient-locator Locator access audit (masked)
RPL-AUDIT-8100
restricted
audit metadata only; does not reveal whether a record exists
asset asset-management Cold-chain unit (synthetic)
AST-0007
operational
program program-performance Malaria program (synthetic)
PRG-MAL
operational
indicator program-performance Suspected cases tested (synthetic)
IND-MAL-02
operational
evidence cross-module Cross-module evidence bundle (synthetic)
EVD-XM-0015
operational
anomaly-case ai-intelligence Cross-module contradiction case (synthetic)
AIC-0015
operational

Cases by source module

  • fund-accountability 3
  • stockpile-logistics 4
  • lab-operations 2
  • patient-continuity 2
  • restricted-patient-locator 1
  • asset-management 2
  • program-performance 2
  • cross-module 1

Confidence distribution

  • confidence 0.0-0.4 0
  • confidence 0.4-0.6 2
  • confidence 0.6-0.8 14
  • confidence 0.8-1.0 1

AI posture — assistive only, non-autonomous

AI assists with

  • search authorized operational records
  • reconcile records across modules
  • detect anomalies and contradictions
  • identify missing or conflicting evidence
  • prioritize review queues
  • summarize evidence for authorized reviewers
  • identify suspicious activity patterns
  • suggest likely root causes
  • recommend next review steps
  • calculate confidence and severity signals
  • surface possible duplicate, fabricated, diverted, or manipulated records

AI must never

  • autonomously approve, deny, suspend, punish, accuse, or sanction
  • autonomously certify financial, clinical, laboratory, asset, or program results
  • autonomously disclose patient identity or precise location
  • autonomously change source records
  • autonomously transfer funds, commodities, assets, or custody
  • autonomously close investigations or compliance cases
  • autonomously suppress findings
  • bypass consent, privacy, role-based access, evidence, separation of duties, or human review
  • represent an anomaly as confirmed misconduct without authorized investigation and decision

AI unifies and prioritizes cross-module signals to help authorized humans review faster. AI confidence is not proof and an anomaly is never confirmed misconduct without authorized investigation and decision. AI may search authorized operational records, reconcile records across modules, detect anomalies and contradictions, identify missing or conflicting evidence, prioritize review queues, summarize evidence for authorized reviewers, identify suspicious activity patterns, suggest likely root causes, recommend next review steps, calculate confidence and severity signals, and surface possible duplicate, fabricated, diverted, or manipulated records. AI never approves, denies, suspends, punishes, accuses, or sanctions; never certifies financial, clinical, laboratory, asset, or program results; never discloses patient identity or precise location; never changes source records; never transfers funds, commodities, assets, or custody; never closes investigations or compliance cases; never suppresses findings; and never bypasses consent, privacy, role-based access, evidence, separation of duties, or human review.

Runtime boundary & linkage