Case Study: Building an AI-Powered Tax Workflow Assistant
I worked with a tax and financial professional to turn a highly manual document workflow into a desktop AI assistant built for real-world IRS case prep.
The process involved pulling information from client intake forms, IRS transcripts, pay stubs, supporting expense documents, and standards tables, then combining it all into a working financial spreadsheet. Before the tool, that meant a lot of repetitive review, cross-checking, and manual data entry.
I built Mac/Windows desktop application that in step 1:
- ingests case documents,
- extracts the relevant data,
- compares values across sources,
- flags discrepancies
- applies IRS standards logic
- and populates the working spreadsheet automatically.
The system uses deterministic rules first and calls the OpenAI API only when a document is messy or unclear, which keeps the workflow more transparent, accurate, and cost-conscious.
Step two is human review and upon approval it converts the spreadsheet into an IRS form 433 pdf for submission to the IRS on client’s behalf.

