The Break Investigator for Regulated Money Movement
Alfred is the operational copilot that investigates what deterministic matching could not close. It explains breaks across FX slippage, interchange deductions, and cutoff timing, with citations to the raw source lines. Read-only: Alfred never moves money.
- Deterministic Matching First
- Raw Line-Item Citations
- Read-Only, Non-Custodial
Root cause identified: 3 SEPA Instant transactions settled on Saturday with interchange deductions from the acquiring bank (Ref: BAI2-88219).
Alfred Core Capabilities
Built to handle the complexity of modern fintech operations at scale.
Natural Language Querying
Finance and operations teams can instantly ask detailed questions across millions of transactions without complex database queries.
Request AccessAutomated Break Root-Cause Analysis
Alfred analyzes timing variances, processor fee deductions, currency conversion spreads, and gateway metadata to explain breaks.
Request AccessZero Hallucination Proof Citations
Every explanation emitted by Alfred references immutable proof objects, exact BAI2 bank record numbers, and raw gateway transaction identifiers.
Request AccessDeterministic Proof Indexes
Alfred queries strictly against deterministic reconciliation indexes and verified proof packs, never guessing balances.
Request AccessDeterministic Intelligence vs Legacy Rules Engines
Generic AI tools hallucinate numbers, while legacy rules engines break under high volume. Alfred queries verified deterministic proof records.
| Legacy Rules Engines | Alfred Proof Intelligence | |
|---|---|---|
| Matching logic | Static rules requiring manual configuration | Deterministic matching indexes backed by Proof Engine |
| Exception investigation | Every exception requires manual spreadsheet review | Automated root-cause analysis with raw feed citations |
| Data format handling | Fails on format variations or timing differences | Normalizes heterogeneous feeds into standardized schemas |
| Audit traceability | Fragmented notes and manual email logs | Cryptographic proof packs and line-item citations |
| Scale | Requires proportional operations headcount | Sub-200ms querying across millions of transactions |
How Alfred Handles Exceptions
Reconciliation intelligence is about deterministic resolution. Your operations team reviews only what genuinely requires human confirmation.
Mismatch detected
A transaction arrives with a discrepancy: amount, account identifier, or timestamp does not match the expected record.
Alfred checks context
Alfred queries historical settlement patterns, related transactions, and source metadata to isolate the root cause.
Deterministic verification
When deterministic proof rules confirm the match, the exception is resolved, logged with complete audit citations, and documented.
Ambiguous case surfaced with context
For complex multi-rail breaks, Alfred surfaces the exception with direct line-item citations to raw bank and gateway feeds.
See Alfred investigate a break
Deterministic matching runs first. Alfred takes what is left, decomposes the variance, and cites the source lines. It reads your records. It never moves money.
The $4,820.50 variance decomposes into three causes. None of it is missing money.
Variance Decomposition
Source Citations
- po_1Ou4Xk · Stripe payout 2026-03-14 · gross $241,180.00
- MT940 :61: 2026-03-14 · credit $236,359.50
- BAI2 record 16 · value date 2026-03-15 · $503.00
Programmatic Access to Intelligence
Embed Alfred's capabilities directly into your internal tools, Slack bots, or customer-facing dashboards.
Natural Language to SQL
Send raw text queries, get structured data back. No parsing required.
Role-Based Scope
API keys can be scoped to specific ledgers or data sensitivity levels.
Async Webhooks
Trigger complex analysis jobs and receive results via webhook.
curl -X POST https://api.naya.finance/v1/alfred/query \
-H "Authorization: Bearer sk_live_..." \
-H "Content-Type: application/json" \
-d '{
"query": "What is the total volume of international transactions in Q3?",
"filters": '{
"currency": ["USD", "EUR"],
"status": "settled"
},
"output_format": "json"
}'
# Response
{
"data": '{
"total_volume": 452900.50,
"currency": "USD",
"transaction_count": 1240
},
"confidence": 0.99
}
Interactive Alfred Demo
Investigate a reconciliation break and trace it back to the raw source lines.