Distributor · Catalog ops Case study

From 200-page supplier PDFs to a live catalog: an automated price-list pipeline

Distributors live and die by their price lists — and most arrive as messy supplier PDFs that someone retypes by hand. We built a pipeline that reads those PDFs with AI, structures every line, and turns a stack of documents into a clean, matched catalog.

Document AIExtractionOps automation
200-page PDFs to a structured catalog, automatically
At a glance 2 min read · 6 sections
  • · Layout-agnostic
  • · Credit-metered
  • · Privacy-minded

The brief

New supplier price lists landed constantly — as PDFs, in every layout imaginable. Staff retyped them into spreadsheets line by line: slow, expensive, and error-prone. The ask was simple: stop the manual data entry without losing accuracy.

The extraction pipeline

  1. 01 Upload Drop in the supplier PDF(s)
  2. 02 Pre-check Detect scanned vs digital, page count
  3. 03 AI extract Gemini reads pages in parallel batches
  4. 04 Structure Rows → SKU, description, price, units
  5. 05 Match Align to the existing catalog
  6. 06 Export Clean Excel workbook out

Pages are extracted in concurrent batches, so a long catalog finishes in minutes rather than a day of typing.

Before and after

Manual data entry
  • Hours per supplier list
  • Typos priced into quotes
  • Bottlenecked on one person
  • Updates lag the market
AI extraction pipeline
  • Minutes per list
  • Consistent structured output
  • Self-serve upload + download
  • Catalog stays current

Built for real ops

Layout-agnostic

Different suppliers, different templates — the AI reads the content, not a fixed schema, so new formats don't need new code.

Credit-metered

Usage runs on a credit balance with an audit trail, so cost per job is transparent and controllable.

Privacy-minded

Uploaded PDFs and generated workbooks are processed and cleaned up — scratch files don't linger in storage.

Self-serve

A simple web app: upload, watch the job, download the workbook. No engineer in the loop for day-to-day runs.

Stack

Extraction
Google GeminiBatched page processingPDF parsing
Backend
FastAPINeon PostgresBackground jobs
Frontend
React + ViteUpload/download flowCredits + audit

The result

The team stopped retyping price lists. Supplier PDFs now go in one side and a clean, matched catalog comes out the other — fast enough that pricing keeps pace with suppliers instead of trailing a week behind.

Frequently asked

Both digital and scanned supplier price lists, across varied layouts. Because the AI reads content rather than a fixed template, new supplier formats work without custom code.

As a clean, structured Excel workbook matched against your existing catalog — ready to import or quote from.

Extraction is structured and reviewable, and the pipeline is built so a person can spot-check output before it feeds quotes — far more consistent than line-by-line manual entry.

Still retyping supplier PDFs?

Send us a sample price list. We'll show you the structured catalog it becomes — and what the pipeline would cost you.

Book a Discovery Call