Skip to content

The 5-Desk AI Verification Flow

BPEVeriFlow distributes applicant verification across five structured, sequential desks. Each desk incorporates AI classification scoring backed by manual override controls available to authorised BPE officers.

Pipeline Overview

flowchart LR
    Start([Application\nSubmitted]) --> D1

    D1["📋 Desk 1\nPre-Screening\n5–10 min"]
    D2["🔍 Desk 2\nDocument Review\n8–15 min"]
    D3["💳 Desk 3\nFinance QC\n10–18 min"]
    D4["⚖️ Desk 4\nDue Process\n7–12 min"]
    D5["👤 Desk 5\nBiometric\n3–7 min"]

    D1 --> |"Score ≥ threshold"| D2
    D1 --> |"Score < threshold"| Ex1[["🚨 Exception\nQueue"]]
    D2 --> |"Approved"| D3
    D2 --> |"Flagged"| Ex2[["🚨 Exception\nQueue"]]
    D3 --> |"AML Clear"| D4
    D3 --> |"AML Flag"| Ex3[["🚨 Exception\nQueue"]]
    D4 --> |"Policy Compliant"| D5
    D4 --> |"Non-Compliant"| Ex4[["🚨 Exception\nQueue"]]
    D5 --> |"ZKP Verified"| Done([✅ Payment\nGateway])
    D5 --> |"Liveness Fail"| Fb["Fallback:\nManual Upload"]
    Fb --> Done
    Ex1 & Ex2 & Ex3 & Ex4 --> |"Officer Resolves"| Resume([Resume Desk])

Desk Summary

Desk Name AI Checks External Integrations Est. Time
1 Pre-Screening OCR form validation, NIN cross-check NIMC Registry 5–10 min
2 Document Review Signature consistency, tampering detection CAC, Immigration Service 8–15 min
3 Finance QC AML/NFIU checks, cash threshold flags BVN, FIRS, NFIU 10–18 min
4 Due Process Privatisation Act compliance, debarment check ICPC, BPE Policy DB 7–12 min
5 Biometric Sign-off Liveness challenges, deepfake detection NIMC Biometric Registry (ZKP) 3–7 min

Total estimated processing time: 33–62 minutes (excluding exception resolution wait times)

Desk 1: Pre-Screening

Objective: Validate form integrity and establish the application parameters.

Actions:

  • Ingests basic applicant details: Full Name, Date of Birth, Address, NIN, BVN, CAC RC Number, and Purpose of Application (Privatisation / Concession / Partnership).
  • Triggers automated OCR analysis on uploaded ID documents, verifying text alignment with entered form values.

AI Engine Scoring: Computes an eligibility likelihood score. Missing fields or OCR mismatches trigger flag notices and may escalate to the exception queue.

Context Trigger: Fires the INIT_APPLICATION reducer, generating a unique Application ID (e.g., BPE-X7K2M), computing the 5-day SLA deadline, and activating Desk 1.

Desk 2: Document Review

Objective: Verify corporate legal existence and director identities against sovereign databases.

Database Integrations:

Database Check Performed
NIMC NIN record lookup to authenticate director biographics
CAC RC number cross-reference to confirm active company status
Immigration Service Director passport validation

AI Engine Scoring: Evaluates Memorandums of Association and Board Resolutions for tampering, signature consistency anomalies, and deepfake digital artefacts.

Desk 3: Finance QC

Objective: Assess solvency and evaluate financial integrity for AML compliance.

Ingested Documents: 12-month bank statements, 3-year audited financial records, FIRS tax clearance certificates.

Compliance Checks (AML/NFIU):

  • Flags cash transactions exceeding the ₦5,000,000 threshold (NFIU regulation).
  • Analyses balance volatility, circular round-tripping transfer anomalies, and debt-to-equity ratios.

AI Engine Scoring: Generates financial health ratings — Low Risk / Moderate Risk / High Risk.

Desk 4: Due Process Enforcement

Known SLA Bottleneck

The Executive Command Center's live metrics consistently identify Desk 4 as the primary SLA bottleneck, with a typical backlog of 40–60 applications. Desk 4 requires multi-level human vetting chains that cannot be fully automated. Executives should monitor this queue closely and pre-assign experienced Compliance Officers to reduce wait times.

Objective: Enforce compliance with the BPE Privatisation Act and ICPC guidelines.

Checklist Auditing:

  • Minimum 3 years operational company history
  • Debarment registry check (no federal blacklisting)
  • Criminal indictment and conflict-of-interest declarations
  • Alignment with the Companies and Allied Matters Act (CAMA) 2020

Multi-Level Approval Chain:

  1. Verification Officer — Initial vetting
  2. Senior Compliance Officer — Secondary vetting
  3. Legal Unit — Action review
  4. BPE Department Director — Final department approval

Desk 5: Biometric Sign-off

Objective: Confirm live physical identity presence of applicant directors via webcam.

Liveness Challenges: Five randomised prompts presented in sequence:

  1. Look straight into the camera
  2. Turn head slowly to the left
  3. Turn head slowly to the right
  4. Blink twice
  5. Smile naturally

Anomalies Checked: Deepfake expression mappings, synthetic silicone masks, and presentation attack vectors.

ZKP Processing: The facial capture is processed locally on-device. Raw video frames are never transmitted — only an anonymised mathematical vector embedding is sent for NIMC registry comparison via a Zero-Knowledge Proof hash.

Fallback Path: If camera hardware fails, the system triggers setBiometricFallback(true), prompting a fallback flow requiring manual physical ID uploads and utility proof-of-address verification.