Audit & Compliance

Anomaly Detection Agent

Scans AP, AR, the general ledger and intercompany for unusual patterns.

All agents Next: Tax & Compliance

An example conversation with the agent

AI Impact

50%

More fraud detected

33%

Fewer duplicate payments

20%

Faster error resolution

Key Capabilities

Transaction Scanning

Scans GL entries for duplicates, outliers, and unusual patterns

Expense Screening

Checks claims against policy thresholds and detects split transactions

Vendor Profiling

Monitors payment frequency, escalation, and new vendor flags

Alert Generation

Produces severity-ranked alerts with supporting evidence

Scenario

Karen Lim, Internal Audit Manager

Reviewing 42,000 transactions for anomalies and control violations

42,000 November transactions scanned — 6 flags across 3 confidence levels

“Scan all November transactions for anomalies — what did you find?”

Baowu Steel INV-88206 — S$445K duplicate, same PO. Third in 12 months: April, August, November

“Show me the Baowu duplicate pattern — is this data entry or something systemic?”

Vendor 8832, Johor Bahru — S$123K, no approval trail, bank in KL but address in Johor

“Investigate Vendor 8832 — run full verification before releasing payment”

JE-9016 — S$1.6M manual entry, no supporting docs, same user created and approved

“What are the segregation-of-duties issues with JE-9016?”

3 lower-confidence flags: ¥15M FX reval (confirmed), Thai Precision overdue (actioned), Kerry Logistics volume (seasonal)

“Which flags are real risks and which are false positives?”

General Ledger Transactions — 1,245,800 Records

Complete journal entry and sub-ledger transaction records across all entities and cost centers

Field	Description
Transaction ID	Unique identifier for each ledger entry
Account Code	GL account number (e.g., 4100-Revenue, 5200-COGS)
Amount	Debit or credit amount in local and reporting currency
Posting Date	Date and time the entry was recorded in the ledger

Entry Types

Type	Description
Automated	System-generated entries from sub-ledgers (AP, AR, Payroll)
Manual	User-initiated adjustments and reclassifications
Reversing	Entries that reverse a prior period accrual or correction

Expense Reports — 87,430 Records

Employee-submitted expense claims with receipts, approvals, and policy compliance flags

Field	Description
Employee ID	Submitting employee identifier and department
Expense Category	Travel, Meals, Lodging, Transportation, Supplies
Claim Amount	Submitted amount with currency and receipt attachment
Approval Status	Pending, Approved, Rejected, Under Review

Vendor Payment Records — 342,150 Records

Accounts payable disbursements to vendors with invoice matching and payment method details

Field	Description
Vendor ID	Unique vendor identifier with master data reference
Invoice Number	Matched invoice reference and PO linkage
Payment Amount	Disbursement amount, currency, and payment date
Payment Method	Wire, ACH, Check, Virtual Card

Audit Trail Logs — 2,180,600 Records

System-level audit trail capturing user actions, data modifications, and access events across financial systems

Field	Description
User ID	Authenticated user performing the action
Action Type	Create, Update, Delete, Approve, Post, Reverse
Timestamp	UTC date and time of the action
Affected Record	Reference to the modified transaction or master data record

scan_transactions

Scan general ledger transactions for unusual patterns, duplicate entries, and statistical outliers across accounts and periods.

Input

account_code date_range entity threshold

Output

Flagged transactions with anomaly scores, pattern classifications, and supporting evidence

detect_outliers

Apply statistical and ML-based outlier detection to identify transactions that deviate significantly from historical baselines.

Input

dataset method sensitivity baseline_period

Output

Outlier records with z-scores, deviation magnitude, and historical comparison

flag_expenses

Screen employee expense claims against corporate policy, detect duplicates, split transactions, and policy violations.

Input

employee_id department date_range policy_rules

Output

Flagged expense items with violation type, risk score, and recommended action

analyze_patterns

Identify recurring behavioral patterns in vendor payments, including frequency spikes, amount escalation, and timing anomalies.

Input

vendor_id lookback_months pattern_type

Output

Pattern analysis with trend visualization, frequency distributions, and risk indicators

search_audit_logs

Query audit trail records by user, action type, time window, and affected data to support investigation workflows.

Input

user_id action_type date_range record_ref

Output

Filtered audit log entries with user actions, timestamps, and change details

generate_alerts

Produce prioritized anomaly alerts with severity ratings, consolidate findings, and route to appropriate reviewers.

Input

findings severity_threshold recipient_group

Output

Structured alert report with severity ranking, evidence summary, and recommended actions

Anomaly DetectionMeridian Industries · November close

Example