Case Study · Data Product

Property Management Reporter

PDF → AI → Tax-Ready Report

I spent 1 hour every month manually reconciling rental statements from 3 different property managers. Now I drag a PDF into a browser and the report is done in under a minute — ready for my accountant at tax time.

← All projects

The Problem

I own 6 rental properties managed by 3 different companies. Every month, 3 PDF statements arrive — each formatted differently, each using different category names, each requiring manual reconciliation before anything useful can be done with the data.

The real problem wasn't just my time. It was that my accountant needed IRS-standard categories and I needed cash-flow visibility. Two users. Two completely different outputs. One messy manual process serving neither of them well.

Users & What I Learned

Primary User
Me — property owner

Wants monthly net income per property at a glance. Doesn't care about line-item detail. Needs speed — not a 2-hour Saturday job.

Secondary User
My accountant

Needs every expense mapped to one of 15 IRS categories. Consistent format across all 6 properties. Totals she can verify without calling me.

Key insight from user research: The accountant is the end user of my output. Designing for her needs first — IRS categories, professional PDF, verifiable totals — made the product actually useful, not just faster for me.

Decisions I Made and Why

PDF output over spreadsheet

My accountant works from PDFs — not Excel. Building a spreadsheet export would've optimized for my comfort, not her workflow. A structured PDF meant she could use it immediately, no conversion step needed.

3-agent pipeline over a single prompt

A single monolithic prompt produced hallucinated totals I couldn't catch. Splitting into extract → normalize → report meant I could validate accuracy at each step before anything hit the final output.

IRS categories as the canonical schema

Each property manager used their own terminology. Rather than preserving it, I normalized everything to IRS Schedule E categories. This made every report tax-ready by default and eliminated the back-and-forth with my accountant about what each line item meant.

Outcomes

< 1 min
per statement
(was 1 hr/month manually)
6
Properties processed across
3 statement formats
0
Manual data entry
steps remaining

I download the statement PDF from each property management portal, drag it into the upload page, and the report is generated in under a minute. At year-end, I hand my tax accountant 6 individual PDFs — one per property, covering all 12 months of revenue and expenses, already mapped to IRS categories. No spreadsheets, no manual data entry, no prep work.

3-Agent Processing Pipeline

Behind every report is an AI pipeline that extracts data, standardises format, and generates professional output.

📋
PDF Upload Statement input
1
Extract Parse PDF
2
Normalize Standardise data
3
Report Generate PDF
Pipeline: Upload statement → Agent 1 parses PDF and extracts transactions → Agent 2 normalises to standardised JSON with IRS categories → Agent 3 generates professional tax-ready PDF with monthly breakdown and totals.

Try the Demo

Select a property to see a preview that matches the actual PDF output — same layout, same data structure. Property names are changed; numbers are representative.

Supported Statement Formats

🏢 HomeRiver Group

Multi-property statements with transaction-level detail.

Format: Property header + transaction rows (income/expenses)

🏠 Fox Rental Management

Single-property statements with summary format.

Format: Month-by-month columns, category rows

📋 Bailey-Foristell

Multi-property summary with categorised expenses.

Format: Property section + expense breakdown

15 IRS Schedule E Expense Categories

1. Advertising
2. Auto & Travel
3. Cleaning & Maintenance
4. Commissions
5. Insurance
6. Legal & Professional
7. Management Fees
8. Mortgage Interest
9. Other Interest
10. Repairs
11. Supplies
12. Taxes & Licenses
13. Utilities
14. Depreciation
15. HOA Dues

Why This Demonstrates PM Skills

📊 AI + Data Discipline

A 3-agent pipeline automates the messy parts (parsing, extraction, categorisation). But the system is only valuable because it solves a real user problem with rigorous data validation.

👥 Secondary User Research

Primary user: property owner. Secondary user: accountant. Both need different things — surfaced through actual interviews, not assumptions.

✅ Quality Assurance

100% accuracy validation against source data. Not just "it works" — verified and tested on 6 real properties across 3 statement formats.