Automate Forensic Accounting Document Ingestion and Reconciliation
Budget: $50 – $0 USD
I need an applied-AI and automation specialist who can look at my existing forensic accounting workflow, currently driven mostly in Excel alongside Word and Google Sheets—and build a repeatable, transparent pipeline that does the heavy lifting while still letting me apply professional judgment before anything goes out the door.
Here is what happens today: I send small to mid-size companies a standardized list of financial and inventory reports that I need to perform an independent bank audit against their loan(s). These reports include year end, tax docs, inventory, insurance, banking docs, financial reconciliation reports, etc. I then upload all paperwork in BOX and reorganize everything for the next phase. Once captured, those figures get reconciled against a set of templated workpapers in Excel, and finally I draft a bank-ready narrative in Word. It works, but it is slow and too manual.
Most of the documents I receive are fairly similar and I can require companies to upload their documents in the manner I see fit. Even though the reports are similar, their naming designations, format and details can very dramatically.
The outcome I’m after is an end-to-end solution that can:
• automatically ingest those source documents,
• extract the key data points with high accuracy, and be able to manually modify to fit the process when necessary,
• populate the existing reconciliation workpapers inside Excel or, where it makes sense, Google Sheets,
• surface any anomalies for my review, not hide them in a “black box”, and
• generate a first-pass narrative that I can edit before sending to banks
I’m open on the tooling—Python, Power Query, VBA, Office Scripts, Google Apps Script, or even LLM-based extractors—as long as the logic remains auditable and easy for an accountant to maintain.
Deliverables
1. A working prototype integrated with my current files and folder structure.
2. Clean, commented source code or scripts.
3. A brief user guide showing how to add new documents, rerun the process and override results when professional judgment calls for it.
4. One live hand-off session (recorded) to walk through the build and answer questions.
If this sounds like your wheelhouse, tell me briefly how you would approach the ingestion and reconciliation stages and what tech stack you’d lean on.
Here is what happens today: I send small to mid-size companies a standardized list of financial and inventory reports that I need to perform an independent bank audit against their loan(s). These reports include year end, tax docs, inventory, insurance, banking docs, financial reconciliation reports, etc. I then upload all paperwork in BOX and reorganize everything for the next phase. Once captured, those figures get reconciled against a set of templated workpapers in Excel, and finally I draft a bank-ready narrative in Word. It works, but it is slow and too manual.
Most of the documents I receive are fairly similar and I can require companies to upload their documents in the manner I see fit. Even though the reports are similar, their naming designations, format and details can very dramatically.
The outcome I’m after is an end-to-end solution that can:
• automatically ingest those source documents,
• extract the key data points with high accuracy, and be able to manually modify to fit the process when necessary,
• populate the existing reconciliation workpapers inside Excel or, where it makes sense, Google Sheets,
• surface any anomalies for my review, not hide them in a “black box”, and
• generate a first-pass narrative that I can edit before sending to banks
I’m open on the tooling—Python, Power Query, VBA, Office Scripts, Google Apps Script, or even LLM-based extractors—as long as the logic remains auditable and easy for an accountant to maintain.
Deliverables
1. A working prototype integrated with my current files and folder structure.
2. Clean, commented source code or scripts.
3. A brief user guide showing how to add new documents, rerun the process and override results when professional judgment calls for it.
4. One live hand-off session (recorded) to walk through the build and answer questions.
If this sounds like your wheelhouse, tell me briefly how you would approach the ingestion and reconciliation stages and what tech stack you’d lean on.