IDP (automatic reading of invoices, Carta Porte, and POD) is not a folder of PDFs. It is reading those documents, matching them to the trip ID, and holding pay if they do not square. If it only extracts text, you bought capture, not Finance.

PDF, XML, photo
Read
rate + CFDI + CP + POD
Match
pattern at 100% audit
5–7%

Cluster: AI automation · use cases · 3PL case.

What IDP is on a freight trip

IDP (automatic reading of invoices, Carta Porte, and POD) means the document enters the trip file, not a folder. Intelligent document processing, on this yard, is not “digitize the folder.” The click / judgment / ticket layer is in AI automation. This page is the document job.

Monday a WhatsApp PDF, a Carta Porte XML, and a CFDI (electronic tax invoice) disagree on weight, origin, or rate. The office scanner extracts letters. Finance still releases pay. That is what the generic “IDP” SERP does not write.

Read, match, hold

Read

PDF, XML, photo

Match

Rate + CFDI + Carta Porte + POD

Hold

Ticket if it does not square

IDP is not a folder of PDFs. It is a document that enters the trip file.

The match playbook (do not clone it here) is in CFDI + Carta Porte + GPS + POD audit and in freight accounts payable.

OCR is not a trip file

LayerWhat it leavesTest
OCRExtracted textDoes it enter the ID?
IDPFields + matchDoes it square with rate and POD?
AuditPay or holdIs there a ticket and an owner?
Three layers. An RFP that asks for “document AI” often delivers only the first.

OCR (optical character recognition) reads. IDP decides. Audit closes or opens a ticket. If the RFP heading says “document AI” and the first sentence does not define those three layers, Monday you bought a scanner with a logo.

Documents that hurt on Mexico–U.S.

Warehouse aisles: the freight document is tested on the trip file, not in a folder
A dirty POD, Carta Porte XML, and a CFDI that does not square. That is corridor IDP, not an office scanner.
  • Invoice vs contracted rate. Case: 2,250 / $3.6M / 5.7%.
  • Carta Porte 3.1 vs what actually rolled (GPS, dock).
  • POD as a WhatsApp photo tied to the trip, not to a chat.

A corridor 3PL does not lose 5–7% in a “vision model.” It loses it when the dock photo never enters the ID and Finance pays invented detention. Published case: 2,250 invoices, $3.6M, 5.7%. What opens the ticket is not the dirty PDF: it is the match that does not square.

Bajío–Laredo scene: the POD photo arrives on WhatsApp, the Carta Porte 3.1 XML states another weight, and the CFDI does not match the contracted rate. OCR reads all three. IDP ties them to the same ID. Finance holds and opens a ticket. What the “intelligent document processing” SERP does not say: an office scanner does not recover 5–7%. The unpublished number: invoices paid Monday without a POD. That opens how to choose.

How to choose

Elige un paso para ver el detalle

Detalle del paso · 01

Does it write the ID?

If the PDF lives on a drive, it is fake digitization.
Minimum so you do not buy a scanner with an AI logo.

Start with the document that hurts: invoice or POD. One corridor. Coexist with the TMS. If the partner does not open a ticket when the XML does not match GPS, you bought capture. OCL can stamp the invoice and Carta Porte when the fiscal flow asks. The buyer owns exceptions.

Ranking of documents, not of scanners: first invoice vs rate, then Carta Porte vs GPS, last the WhatsApp POD. A drive of PDFs is fake digitization. The product is pay or hold. A 6–8 week pilot. One document type. One corridor.

CFDI (Mexico’s digital tax invoice) without a match is capture. Carta Porte without GPS is a file, not a trip. The CEO question Monday: how many invoices left without a POD on the same ID.

Related reading

Key takeaways5 points
  1. Freight IDP = read + match + hold. Not a folder of PDFs.
  2. OCR extracts. The file decides pay or not.
  3. Four sources: rate, CFDI, Carta Porte, POD.
  4. OCL pattern at 100%: 5–7%. Case 2,250 / $3.6M / 5.7%.
  5. 6–8 week pilot. OCL can stamp when the fiscal flow asks.

Does the PDF arrive and Finance still pays blind?

In 30 minutes we look at one document type and whether the match opens a ticket before pay.

Frequently asked questions