FinClose AI – Agentic Financial Close Automation

Workflow Diagram
Architecture Diagram

Inspiration

Financial close is one of the most critical yet time-consuming processes in every enterprise. Finance teams spend days manually reconciling data across ERP systems, investigating discrepancies, validating evidence, and obtaining approvals before books can be closed.

We wanted to build an intelligent solution that combines AI agents, automation, and human oversight to dramatically reduce the effort required for financial close while improving auditability and compliance.

What it does

FinClose AI is an Agentic AI platform built on UiPath Maestro BPMN that orchestrates multiple specialized AI agents to automate the financial close lifecycle.

The solution:

Performs automated reconciliation across financial datasets
Detects and prioritizes financial exceptions
Performs Root Cause Analysis (RCA)
Routes exceptions to specialized Finance, Treasury, or Tax agents
Collects supporting evidence automatically
Generates audit-ready recommendations
Sends final decisions for human approval using UiPath Action Center

The platform enables organizations to move from manual investigations to AI-assisted financial operations with complete governance and human-in-the-loop control.

How we built it

The solution is designed as a multi-agent orchestration workflow using UiPath Maestro BPMN.

Reconciliation Agent

Reads financial datasets
Performs reconciliation
Detects exceptions
Generates reconciliation summary

Root Cause Analysis Agent

Investigates each exception
Identifies probable cause
Determines responsible business function

RPA Workflow

Retrieves supporting ERP information
Stores investigation artifacts
Prepares structured evidence

Domain Specialist Agents

Finance Agent

General Ledger validation
Journal Entry analysis
Financial Close Task validation

Treasury Agent

Bank reconciliation
Cash position validation
Payment investigation

Tax Agent

GST/VAT compliance
Tax reconciliation
Filing validation

Merge Gateway

All specialist agent outputs are synchronized before proceeding to audit.

Audit Agent

The Audit Agent:

Validates supporting evidence
Verifies compliance rules
Creates audit-ready recommendations
Determines financial close readiness

Human-in-the-Loop

High-impact recommendations are routed to UiPath Action Center where finance users can:

Review evidence
Approve recommendations
Reject recommendations
Add comments

This ensures AI recommendations always remain under human governance.

Challenges we ran into

During development we encountered several challenges:

Orchestrating multiple AI agents using UiPath Maestro
Designing structured communication between agents
Managing large financial datasets efficiently
Creating robust exception routing logic
Implementing human approval using UiPath Action Center
Handling missing financial datasets gracefully during reconciliation
Maintaining auditability across every workflow stage

We also learned how important structured outputs are when working with Agent Builder and Pydantic models, especially ensuring agent responses strictly follow expected schemas.