BAEZ TECH SOLUTIONS

Overview

A lending operations team needed a faster, more reliable way to review customer bank statements as part of their loan origination process. Instead of manually reading multi-page OCR exports and rebuilding summaries in spreadsheets, we implemented an OpenAI-powered financial auditor that ingests raw OCR text from bank statements stored in Google Drive and produces a clean, consolidated monthly summary.

This case study outlines how we orchestrated the workflow with Monday.com, Jotform, Google Drive, and OpenAI to turn messy statement text into decision-ready financial insights.

Goal / Objective

• Automate the review of bank statements as part of a loan origination (LOS) workflow.
• Transform unstructured OCR text from bank statements into structured, analyzable data.
• Identify distinct monthly statement periods across multiple pages and accounts.
• Extract key financial figures from each statement’s Summary section.
• Produce a consolidated report that underwriters and financial analysts can use immediately.

Challenge / Problem

Traditional bank statement review in lending workflows has several pain points:

• Unstructured OCR text: Bank statements are often scanned and converted to text with inconsistent formatting, line breaks, and artifacts.
• Multiple pages and periods: A single upload may contain several months of statements, sometimes across different accounts or banks.
• Manual consolidation: Analysts must scroll through pages of text, locate each “Summary” section, and manually copy figures into spreadsheets.
• Risk of human error: Misreading a number or missing a month can materially impact underwriting decisions.

The client needed a way to reliably interpret this raw OCR text and surface the most important financial figures with minimal human effort.

The Workflow / Solution

We designed an AI-assisted workflow that uses Monday.com as the orchestration layer, Google Drive as the document store, and OpenAI as the analysis engine.

1. Intake and Task Creation

• Bank statements are uploaded and stored in Google Drive.
• A new item is created in the Portfolio / LOS workflow (either directly in Monday.com or via a Jotform intake form feeding into Monday).
• The task includes links to the relevant Google Drive files and the raw OCR text output from the statement processing step.

2. Triggering the Financial Audit

• When a new “Financial Audit” task is created or moved into the audit stage in Monday.com, an automation kicks off.
• The automation collects the unstructured OCR text for all uploaded bank statement pages associated with that task.
• The text payload is sent to an OpenAI financial auditor prompt tuned specifically for bank statement analysis.

3. AI Analysis of Bank Statements

The OpenAI workflow is instructed to:

1. Identify distinct statement periods
   • Detect start and end dates for each monthly statement.
   • Group pages belonging to the same month or statement period.
2. Locate each “Summary” section
   • Scan the OCR text for sections labeled “Summary”, “Account Summary”, or similar variants.
   • Isolate the lines inside those sections even when formatting is inconsistent.
3. Extract key financial figures (when present), such as:
   • Opening balance
   • Closing balance
   • Total deposits / credits
   • Total withdrawals / debits
   • Fees assessed
   • Average balance or minimum balance
4. Normalize and validate values
   • Standardize currency formats and negative/positive signs.
   • Cross-check that opening/closing balances align with total debits and credits when possible.
5. Generate a consolidated report
   • Output a structured, human-readable summary for each month.
   • Provide a consolidated view across all months included in the upload.

4. Consolidated Output Back to the LOS

• The AI’s structured output is returned in a predictable JSON-like structure and formatted into a clear, readable report.
• This report is attached back to the task (in Monday.com and/or ClickUp) as:
  • A formatted summary section in the task, and/or
  • A linked document that underwriters can reference.
• Optional flags or notes are added for:
  • Missing or inconsistent months.
  • Large swings in balances or unusual transaction patterns.
  • Any parsing ambiguities that may require human review.

Results / Impact

While this implementation is primarily focused on workflow and data quality rather than marketing metrics, the solution is designed to deliver:

• Dramatically reduced manual review time: Analysts receive a ready-made monthly breakdown instead of scrolling through raw OCR pages.
• Higher consistency and fewer errors: The same extraction logic is applied every time, reducing the chance of missed months or mis-keyed values.
• Faster underwriting decisions: With clean monthly summaries, lending or risk teams can focus on judgment calls rather than transcription.
• Scalability: The pipeline can handle many concurrent statement reviews without requiring additional headcount.

Tools Used

• Monday.com – Orchestrates the LOS workflow, tracks each audit request, and stores AI outputs alongside tasks.
• Jotform – (Optional) Used to collect client inputs or upload instructions that feed into Monday.com.
• Google Drive – Secure storage for uploaded bank statement files and OCR outputs.
• OpenAI – Core financial auditor engine that interprets OCR text, extracts figures, and builds the consolidated report.
• Make.com – Automates integrations and workflows between tools, enhancing process efficiency.