Banking Intelligence

AI that a bank board can defend.

VoxID unifies AML monitoring, salami-slicing fraud detection and an agentic AI contact centre on one auditable platform — with live, Power BI–class intelligence for the C-suite and the regulator.

AML & CFTSalami-slicingAgentic Contact CentreVoice BiometricsAudit-ready
banking ops · automation
Live VoxID console
Self-service
71%
Loan inquiries
3,402
Dormant reactivated
612
STP rate0.92
STP rate. 10 points, trend up, min 0.20, max 0.92, average 0.55, latest 0.92.
  • STP rate0.92
+82%
AML alerts triaged
-46%
False positives
8s
Avg verify time
-37%
Contact-centre AHT
AML & CFT Monitoring

From rule-based alerts to explainable AI typologies.

Traditional AML systems drown analysts in noise — 90%+ false-positive rates are not unusual. VoxID layers behavioural baselining, network analytics and voice-biometric verification on top of your existing rules engine, so every alert arrives with a rationale, a risk score and a recommended next action.

  • Structuring & smurfing detection
  • Round-trip & layering graphs
  • PEP / sanctions screening
  • Beneficial-owner risk scoring
  • Trade-based ML indicators
  • Mule-account clustering
  • STR/SAR auto-drafting
  • Regulator-ready audit trails
fraud + aml · risk board
Risk live
Fraud alerts
27
AML cases
342
Saved (MTD)
KES 2.4B
Suspicious tx velocity0.46
Suspicious tx velocity. 10 points, trend up, min 0.30, max 0.70, average 0.44, latest 0.46.
  • Suspicious tx velocity0.46
Composite risk
Power BI · Transaction Monitoring

The AML command centre.

Switch perspectives the way a Tier-1 analyst does — from the global alert queue, into a single typology, all the way down to the case file.

aml monitor · live
Drill-down
Open alerts
1,284
-18% WoW
Avg triage
4m 12s
-62%
True positive
37.4%
+21pt
STRs filed
118
+9
High-priority queue
  • STRUCTAcct 4471·22 · 14 deposits ≤ KES 99,5000.91
  • PEPWire to sanctioned BO match (87%)0.87
  • MULECluster of 23 fan-in beneficiaries0.82
  • TBMLInvoice over-valuation · LC #884120.74
fraud + aml · risk board
Risk live
Fraud alerts
27
AML cases
342
Saved (MTD)
KES 2.4B
Suspicious tx velocity0.46
Suspicious tx velocity. 10 points, trend up, min 0.30, max 0.70, average 0.44, latest 0.46.
  • Suspicious tx velocity0.46
Composite risk
Salami-Slicing Fraud

Catch the invisible thousand cuts.

Salami-slicing attacks siphon tiny amounts — fractions of a cent on interest postings, rounding remainders on FX conversions, micro-debits across millions of dormant accounts — that individually fall below every alerting threshold. VoxID's anomaly engine watches the aggregate behaviour of beneficiary accounts, GL suspense buckets and rounding journals to surface the pattern before it becomes a board-level incident.

Rounding-journal monitor
Flags GL accounts accumulating sub-cent rounding gains.
Beneficiary graph
Clusters many-to-one micro-credit fan-ins across channels.
Velocity baselining
Per-account, per-hour spend signatures with seasonal drift.
Forensic replay
Cycle-accurate replay of a suspected operator's session.
How detection works

Three signals turn noise into evidence.

Each slice is invisible; the pattern is not. VoxID stacks behavioural, network and ledger evidence and only escalates when all three converge.

1 · Behavioural baseline
Per-account spend rhythm, hour-of-day, FX corridor. Drift detector flags > 3σ deviation in any window.
2 · Beneficiary graph
Build a directed graph of micro-credits. Fan-in > 50 unique source accounts to one BO over 30 days raises a cluster alert.
3 · Ledger reconciliation
Reconcile GL suspense, rounding and interest journals nightly. Sub-cent accumulations are auto-quarantined.
Before VoxID · 12-month loss exposure
Live VoxID console
  • Loss trend+406% YoY
Detection
9%
MTTD
94d
Loss
KES 608M
After VoxID · 12-month risk reduction
-94%
  • Loss trend-94% YoY
Detection
96%
MTTD
2.4h
Loss
KES 38M
Agentic AI Contact Centre

Agents that act, not just answer.

A VoxID agent authenticates the caller by voice in under 8 seconds, retrieves the right context from your core banking and CRM, executes the transaction with policy-aware tool use, and hands a clean summary to a human only when the situation truly requires it. The result: fewer security questions, lower AHT and a measurable lift in first-contact resolution.

Voice-first auth
Frictionless biometric login on every channel.
Tool-using agents
Calls core banking APIs with guardrails & approvals.
Agent assist
Live next-best-action and call summarisation.
Real-time QA
Every call scored on compliance, sentiment, resolution.
contact centre · live ops
Live VoxID console
In queue
14
Avg AHT
1m 42s
Auto resolved
64%
Agentic resolutions
  • VOICEVerified · balance shared7s
  • AGENTStop-payment executed11s
  • ESCALATEHigh-risk → humanreview
Walkthrough

A call, end-to-end, in under 30 seconds.

Follow the same call your floor manager hears every day — from ring to resolution — and watch the Power BI deflection chart fill in real time.

  1. 1
    0s · Ring
    Caller routed; voiceprint capture begins on the first 'hello'.
  2. 2
    7s · Auth
    Voice + device + behavioural score → silent pass. No security questions.
  3. 3
    11s · Intent
    Agentic NLU classifies: 'stop payment on card ending 4471'.
  4. 4
    16s · Tool use
    Agent calls core banking API; policy guardrail confirms limit & dual-control.
  5. 5
    22s · Action
    Card blocked, replacement ordered, dispute case opened, SMS sent.
  6. 6
    28s · Wrap
    Compliance-graded call summary written to CRM; QA score 96/100.
contact centre · outcomes (Power BI)
Live
Deflection
71%
▴ +18pt
Avg resolve
1m 06s
▾ -47%
Compliance
99.2%
▴ +6pt
  • Self-service deflection
  • First-contact resolution
  • Compliance score
Resolution rate0.84
Resolution rate. 10 points, trend up, min 0.42, max 0.84, average 0.66, latest 0.84.
  • Resolution rate0.84
Executive Intelligence

A Power BI-class view your CEO will actually open.

Every signal — AML, fraud, contact centre, channel adoption — flows into a single, drill-down executive dashboard with branch, region and product slices.

executive command centre
Boardroom
Verified
184k
Auth
98.7%
Auto
64%
Fraud↓
38%
Resolved by AI / hr0.92
Resolved by AI / hr. 10 points, trend up, min 0.20, max 0.92, average 0.55, latest 0.92.
  • Resolved by AI / hr0.92
Composite KPI
banking ops · automation
Live VoxID console
Self-service
71%
Loan inquiries
3,402
Dormant reactivated
612
STP rate0.92
STP rate. 10 points, trend up, min 0.20, max 0.92, average 0.55, latest 0.92.
  • STP rate0.92
Executive case study

A Tier-1 East African bank, 12 months in.

Anonymised numbers from a production VoxID deployment across 142 branches, 4.1M customers and a 380-seat contact centre.

"VoxID gave us, for the first time, an AML and contact-centre view our board could actually act on. Inside one quarter we cut analyst backlog, killed a salami-slicing pattern that had run for two years, and reduced AHT without losing a single compliance point."

Chief Risk Officer
Tier-1 commercial bank · East Africa
Fraud losses
-KES 570M
AML false positives
-62%
Analyst productivity
+3.4x
Avg handling time
-47%
ROI scenario · live

Adjust the assumptions. Watch NPV and payback update.

Bank size
Alert volume
Analyst cost
3-year NPV
USD 24.6M
Payback
1.5 mo
Annual benefit
USD 9.8M
Investment
USD 968k
Modelled: ~31 analyst-FTE redeployed · fraud-loss avoidance · AHT savings. Discount rate 10%.
Compliance · Security · Governance

Proof a regulator and a CISO can both sign.

Every claim above is backed by documented controls, third-party assurance and a model-governance regime designed for supervised banking environments.

Data handling
ISO 27001 · SOC 2 II

Voiceprints encrypted at rest (AES-256) and in transit (TLS 1.3). Tenant-isolated KMS keys. Data residency on request (KE, ZA, AE, EU).

Auditability
Immutable trail · WORM

Every model decision, prompt, tool call and human override is captured with cryptographic hash chaining. Exportable for regulators in <60s.

Model governance
NIST AI RMF · ISO 42001

Versioned models with documented training data, bias testing, drift monitoring and a four-eyes change-approval workflow.

Regulatory fit
CBK · CMA · GDPR · DPA 2019

Pre-mapped controls for Kenya DPA, GDPR, CBK Risk Mgmt Guidelines, FATF 40 recommendations and Basel operational-risk reporting.

Evidence badges · click for control mapping
Documentation
Control coverage
Access control & RBAC98%
Encryption (rest + transit)100%
Logging & monitoring96%
Model risk management94%
Third-party assurance92%
Independently assessed · last audit Q1 2026
Use cases

Built for every team that touches the customer.

For Banks
  • Voice verified call centre
  • Account balance self service
  • Stop payment automation
  • Loan eligibility checks
  • Dormant account activation
  • Fraud monitoring
  • AML alerts
  • Customer onboarding
  • Diaspora verification
  • High value transaction confirmation
For Saccos
  • Member verification
  • Loan balance inquiry
  • Savings balance inquiry
  • Loan eligibility check
  • Guarantor verification
  • Member support automation
  • Fraud alerts
  • Board level analytics
  • Branch performance dashboard
  • WhatsApp member self service
For Contact Centres
  • Reduce average handling time
  • Reduce security questions
  • Authenticate before agent engagement
  • AI call summaries
  • Agent assist
  • Smart escalation
  • Complaint routing
  • SLA tracking
  • Quality monitoring
For Risk & Compliance
  • AML monitoring
  • Suspicious pattern detection
  • Voice mismatch reports
  • Audit trails
  • Investigation case management
  • Regulatory reporting support
  • Role based access
  • Approval workflow