ARVEXI
ProductCortex

The AI that investigates your accounts

Other platforms flag variances. Cortex investigates them. Querying GL transactions, comparing prior periods, and generating audit-ready work papers with cited evidence. Your team reviews completed analysis instead of starting from scratch.

Cortex Investigation

Account 4210 · Accrued Liabilities

Investigating...
0 tool calls·0.0s·$0.08
Powered byCortex
Cushman & Wakefield
Planet Fitness
Prologis
Waymo
Marcus & Millichap
Acrisure
BlueLinx
Cushman & Wakefield
Planet Fitness
Prologis
Waymo
Marcus & Millichap
Acrisure
BlueLinx

The AI maturity framework

Four levels of AI in accounting

Level 1

Commentator

Flags variances, assigns risk scores, generates alerts. The human still does all the work.

Oracle ARCS, BlackLine

Level 2

Investigator

AI autonomously investigates using structured data tools. Produces findings with specific evidence.

Arvexi Cortex

Level 3

Preparer

AI completes the entire reconciliation end-to-end: creates items, documents findings, submits for review.

Coming soon

Level 4

Controller

AI manages the entire close: orchestrating tasks, resolving dependencies, escalating only material exceptions.

Our vision

Confidence Scoring

Account 4210 · Accrued Liabilities

5-factor deterministic

Step 1. Score

Every account scored with a 5-factor formula

Cortex scores every reconciliation using five deterministic factors. The formula is auditor-reproducible. Same inputs, same score, every time. GREEN means low risk. AMBER means review. RED triggers investigation. Read about confidence scoring.

cortex / investigation agent

Step 2. Investigate

7 data tools, real transactions, structured findings

For amber and red accounts, Cortex launches an autonomous investigation agent. It queries your GL, compares prior periods, checks IC counterparts, and produces findings with specific evidence. Learn how AI investigation agents replace manual reconciliation.

Work Paper, AI Generated

4210 · Accrued Liabilities

AMBER · 83.4

March 2026 Close

Balance Summary

GL Balance

$847,230.00

Subledger Balance

$834,890.00

Unreconciled

$12,340.00

Finding 1

TIMING

Variance of $8,200.00 is attributable to an accrual reversal posted on March 28 that has not yet cleared through the subledger. This pattern matches the prior three periods where reversals cleared within 3 business days of period end. Expected to resolve by April 3.

Cited Transactions

JE-2026-03-1847Accrual reversal, utilities
Mar 28$8,200.00
JE-2026-03-1832Monthly accrual, utilities
Mar 1$8,200.00

Reconciling Items

TTiming: Accrual reversal pending clear
$8,200.00✓ Created
Cortex v3.2·33.1s·7 tool calls
Generated Mar 17, 2026 02:14 AM

Step 3. Document

Formal work papers generated automatically

Cortex produces structured work papers with balance summaries, variance analysis, investigation findings with cited transactions, and auto-created reconciling items. Every finding references specific data from your database. Read about AI agents in accounting.

Decision Pipeline

Action: Create reconciling item RI-4210-0047

Processing...

Layer 1: SOX Boundary

Journal postings, certifications, approvals

Layer 2: Materiality

Amount: $8,200 < threshold $50,000

Layer 3: Confidence

Score: 83.4% > minimum 70%

Step 4. Trust

Three-layer pipeline enforces SOX boundaries

Every Cortex action passes through three layers. SOX-critical actions are permanently blocked. Materiality thresholds control automation. Confidence self-awareness downgrades uncertain actions to suggestions. Follow our Cortex configuration guide.

Cost transparency

$7 per full sweep. Less than your team's coffee budget.

$0.00

Per GREEN account

Template work papers, no AI call

$0.02

Per AMBER/RED narrative

AI-generated analysis

~$7

Full investigation sweep

210 accounts, all findings

7Data tools
47GL transactions queried
33sPer account
$0.08Per full sweep

Gets smarter every cycle

Cortex learns from your controllers

Calibration feedback

Controllers mark scores as correct, too high, too low, or wrong. Each calibration adjusts future scoring for that specific account.

Cross-cycle patterns

Detects recurring variances, false positives, timing differences, and seasonal patterns. Timing items that resolve every month stop being flagged.

Cortex rules

Configurable per-organization rules: threshold overrides, active/inactive toggles, and condition-based logic.

FAQ

Ready to meet your AI senior auditor?

Book a demo