Command Overview
demo_fintech · All transactions today
Total Decisions
—
All time
Allowed
—
Auto-approved
Pending
—
Awaiting approval
Blocked
—
Policy violations
Executed
—
Completed
AI Agent Activity
—
AI decisions observed
Avg Decision Time
—
ms to control outcome
Fraud Detected
—
Flagged by SOVIGL
AI Visibility
AI Agent: —
Last Action: —
Last decision at: —
Decision: —
Reason: —
All times in UTC (UTC)
Pending Approvals
| ID | User | Amount | Type | Status | Reason | Role | Created | Action |
|---|---|---|---|---|---|---|---|---|
| Loading... | ||||||||
All times in UTC (UTC)
Approval Queue
Review and action pending payment decisions
| ID | Decision ID | User | Amount | Type | Role | Status | Reason | Stage | Created | Action |
|---|---|---|---|---|---|---|---|---|---|---|
| Loading... | ||||||||||
All times in UTC (UTC)
What is SOVIGL?
SOVIGL ensures every AI-driven financial decision is controlled, explainable, and auditable before money moves.
Problem
- AI agents can trigger payments autonomously at machine speed.
- Without policy controls, high-risk payments can execute before human review.
- Many systems lack clear decision reasons for each action.
- Audit trails are often incomplete for compliance and post-incident checks.
What SOVIGL Does
ALLOWED
Payment proceeds immediately when it stays inside approved policy boundaries.
PENDING
Payment pauses and waits for explicit human approval when thresholds require it.
BLOCKED
Payment is stopped when policy, mandate, or meta rules are violated.
How It Works
1. Request sent
2. Policy, mandate, and meta_rules checked
3. Decision made (allowed, pending, or blocked)
4. Decision logged with status, reason, and policy version
5. System acts (execute, wait for approval, or stop)
Why It Matters
- Permanent audit logs provide defensible traceability for every decision.
- Policy version tracking shows exactly which controls were active.
- Regulatory explainability is built-in through explicit decision reasons.
What SOVIGL Is / Is Not
Is
- Decision gatekeeper for AI-driven payments
- Policy enforcement layer before execution
- Audit and governance system of record
Is Not
- Payment processor
- AI decision-maker
- Banking core or ledger
Live Decision Feed
Every payment is evaluated before execution
Payment → Decision → Reason
Real-time Payment Decisions
—
Loading feed...
Allowed / Blocked / Pending based on policy rules
All times in UTC (UTC)
All times in UTC (UTC)
Audit Logs
Complete tamper-proof decision trail · Compliance Rec 7
| Timestamp | Event | Decision ID | User | Status | Reason | Policy Ver |
|---|---|---|---|---|---|---|
| Loading logs... | ||||||
All times in UTC (UTC)
Action Type Reconciliation
Detect seen action types without configured policy rules
Seen Action Types
—
From decision logs
Configured Type Rules
—
Policy type_rules keys
Types Without Rules
—
Potential governance gaps
Policy type reconciliation
strict_types: —
Seen Action Types
—
Configured Type Rules
—
Types Without Rules
—
AI Governance Map
Technical controls supporting EU AI Act · MAS FEAT · NIST AI RMF · RBI FREE-AI
🇪🇺 EU AI Act
✔
Art. 9 | Risk classification system
risk_score + risk_level on every decision
✔
Art. 12 | Audit trail for every decision
decision_id + policy_version + timestamp logged
✔
Art. 13 | Explainability of AI decisions
explanation_registry with plain_english on every decision
✔
Art. 14 | Human oversight before execution
pending approval workflow with human-in-loop
🇸🇬 MAS FEAT
✔
Accountability | Who made the decision
user_id + agent_id + role logged on every action
✔
Traceability | Full decision trail
complete audit log with decision_id + timestamps
✔
Transparency | Why the decision was made
reason + plain_english explanation on every decision
✔
Separation of Duties | Role-based approval chain
employee → manager → admin approval workflow
🇺🇸 NIST AI RMF
✔
GOVERN | Policy version controls
policy versioning + mandate controls per org
✔
MEASURE | Quantify AI risk
risk_score 0.0-1.0 + risk_factors on every decision
✔
MANAGE | Respond to risk outcomes
approved / pending / blocked enforced before execution
✔
MONITOR | Ongoing operational oversight
live feed + health endpoint + reconciliation
🇮🇳 RBI FREE-AI
✔
REC 04
Role thresholds enforce allowed/pending/blocked outcomes
✔
REC 07
Audit logs include decision_id, policy_version, reason
✔
REC 11
Human-in-loop stage approvals for high-risk actions
✔
REC 12
Separation of duty via approver role + identity tracking
✔
REC 16
Policy version visibility via API and dashboard
✔
REC 18
Lifecycle control: pending → approved/rejected → executed
✔
REC 21
Model explainability via explanation_registry on every decision (API + audit)
✔
REC 24
Independent risk scoring via risk_assessment on every decision (API + audit)
✔
REC 26
Operational monitoring with live feed + API health
LIVE PROOF — All controls verified against demo_fintech in real time
Live Proof (demo_fintech)
| REC | Control | Proof Status | Live Evidence |
|---|---|---|---|
| Loading proof... | |||
Controls support compliance. Certification requires independent audit. Contact sovigl100@gmail.com for enterprise compliance documentation.
Live Demo
Decide once. Enforce everywhere.
We make one decision. Everyone else simply follows it.
Other systems carry out the payment — they don’t re-check the rules one by one.
Choose scenario
Steps
AI Agent
Intent
Rules check
Sealed
Safety seal
Proof saved
Bank reads
Pay
Decision made in: — ms
⚡ The answer was ready before the bank moved money
Timing is one real check above. The steps below replay so you can read them.
Without SOVIGL: many checks add up · With SOVIGL: one clear answer (— ms).
READY
Decisions
—
Blocked
—
Pending
—
Decision ms
—
Chain (cumulative)
—
Your decision (sealed)
This is the answer everyone shares
- If something changes, a new decision is created
- Same answer at the bank and in the records
- The bank checks this — it doesn’t re-run all the rules
Decision #1 (current)
Future changes → new decision (#2, #3…)
Details
Waiting for decision…
This decision is sealed for safety.
If anything changes, a new decision is created.
Nothing is hidden or overwritten.
If anything changes, a new decision is created.
Nothing is hidden or overwritten.
Same try again? We reuse this decision — no double payment.
Bank
Waiting…
Proof trail
Waiting…
What happened (simple)
Why this matters
Normally, every system checks again and again. Here, the decision is made once and reused everywhere.
Compare the world
✅ One clear answer
✅ Bank and records stay in sync
✅ No mixed messages when money moves
System 1 checks ❓
System 2 checks ❓
System 3 checks ❓
Conflicts ✗ · Fraud can slip ✗ · Slow ✗
Replay with a twist
Transaction Simulator
POST /simulate with audit logs and demo approval rows; server auto-generates decision_id if omitted
MANUAL CHECK
User ID
Role
Amount
Type
Agent ID
Description / Note
Vendor (optional)
Category (optional)
Select role to view policy limits.
AUTO SIMULATE
Transactions
Delay (ms)
Output Console
Console timestamps: UTC (UTC)
// SOVIGL Control Plane ready
> Send a transaction to begin