Get a demo

Customs document digitization

Turn customs entry summaries, commercial invoices, and broker PDFs into structured rows. Pull line-level data from 25-to-300-page documents and push it into your platform automatically.

Get a demo

The prompt

I want to automate customs document parsing. Build me a flow that pulls customs entry summaries, commercial invoices, and broker PDFs from email or a portal, extracts the line-level data including HTS codes, country of origin, and duty amounts, and pushes the structured rows into our platform via API.

Just copy and paste the prompt into a new Parabola flow to get started.
Parabola flow categorizing and reconciling freight invoice charges

What Parabola builds

A workflow with six steps you can edit:

1. Receive the document. Forwarded broker email, uploaded entry summary, or a pull from a broker portal.

2. Extract the line items. The AI step pulls vendor, HTS code, description, quantity, unit value, country of origin, duty rate, and duty paid per line.

3. Validate against the booking. Cross-check each line against the original commercial invoice and the booking record.

4. Apply classification rules. Look up the HTS code, confirm duty rate, surface anything that doesn't match the expected classification.

5. Output to the platform. Push structured rows via API into your platform or your data warehouse.

6. Flag the exceptions. Lines that don't match the booking, duty amounts outside tolerance, missing fields. These land in a review queue.

Why teams stop doing this manually

Customs paperwork is the bottleneck on every international shipment. Entry summaries run 25 to 300 pages. Each line has an HTS code, country of origin, duty rate, and duty amount. The customs team or the ops team keys the data from the PDF into a spreadsheet, then into the platform, then into the data warehouse.

The work is slow and unforgiving. A misclassified HTS code shows up in a duty audit six months later. A missed country-of-origin field can hold a shipment at the port. The volume is high and the tolerance for error is low.

Some teams add headcount. Others give up and lose visibility, coding everything to a generic GL account and accepting that the line-level detail is gone. Neither path is sustainable as volume grows.

The fix is to extract everything mechanically and put the human attention on classification judgement and exceptions. A flow handles the PDF-to-rows part. The customs team reviews exceptions and confirms classifications. The clean lines hit the platform without anyone typing.

How it works

Step 1. Paste the prompt.

Open Parabola, paste the prompt in section 2, and let it ask follow-up questions about your broker mix, your platform's API spec, and your HTS classification logic.

Step 2. Connect your data.

Broker inbox or portal, your platform's API, your HTS reference table, and any internal classification rules.

Step 3. Run it on every shipment.

New entry summary arrives, the flow parses it, validates it, pushes the structured rows into the platform, and flags exceptions. Customs reviews the queue.

Related prompts

FAQ

How long are typical entry summaries you can handle?

25 pages is normal. 300 pages is supported. The AI extraction step handles multi-hundred-page documents without splitting.

Does the flow handle multiple brokers with different document formats?

Yes. Each broker gets a learned pattern. The first few shipments train the flow; after that the parsing is automatic.

Can it validate HTS classification?

Yes. Lookup against your HTS reference table. Anything that doesn't match the expected classification for that product or supplier gets flagged.

What about partial shipments and split entries?

Each entry parses on its own. Cross-shipment reconciliation is a follow-on flow that ties partial entries back to the original booking.

How is this different from a broker portal's report?

The broker portal shows what they filed. The flow turns the filing into structured data inside your systems. Different audience, different fidelity.
Stop keying customs PDFs into spreadsheets.
Paste the prompt, point it at your broker inbox and your platform, and let the line-level extraction run on its own.
Start for free