How can I automatically verify that an invoice and a packing list match?

A commercial invoice lists 12,500 kg.

The packing list lists 12,600 kg.

And the bill of lading lists 12,850 kg.

However, all three documents are perfectly legible. The data was extracted correctly. There were no OCR issues and no apparent reading errors.

So, what's the problem?

The documents do not tell exactly the same story.

This is a common situation in international transportation and trade. A single shipment is documented by several different documents: commercial invoice, packing list, bill of lading, purchase order, customs documents, certificate of origin, and freight invoice. Each document contains some of the information required for the transaction.

The real challenge isn't just being able to read these documents. It's being able to automatically answer a much more important question:

Does the information in these documents refer to the same transaction?

That is where the document reconciliation begins.

Invoice and Packing List: What Information Must Match?

Before discussing automation, we need to distinguish between the roles of these two documents.

These documents contain some common information, but they are not meant to be identical.

An invoice and a packing list must be consistent in the information they have in common; they do not necessarily have to contain the same data.

Key Data Points to Compare

The first principle of automatic verification is to determine which data points actually need to match.

Why Comparing an Invoice and a Packing List Isn't Enough

At first glance, the problem seems simple. All you would have to do is take the information from the invoice and the packing list, and then check whether the values match. But this approach quickly reaches its limits.

A difference isn't necessarily a mistake

Let's imagine:

Invoice: 1,000 units
; Packing list: 960 units

A comparison system immediately detects any discrepancies.

But if the order is for 960 units, the situation is perfectly understandable.

These may include:

• a partial shipment;
• a modified order;
• an invoice that needs to be corrected;
• a difference in scope between the documents.

The system should avoid automatically treating every discrepancy as an error.

He must try to understand the discrepancy.

Comparison or reconciliation: What's the difference?

The comparison raises a simple question:

Are the two values the same?

Reconciliation raises a business-related question:

Do these two pieces of information describe the same reality, and if they differ, why?

This distinction is fundamental.

Document reconciliation involves comparing the information in multiple documents to identify inconsistencies before they become operational problems.

Comparing helps us identify a difference. Reconciling helps us understand that difference.

‍

How can I automate the process of verifying that an invoice matches a packing list?

Reliable verification involves several steps. The goal is not simply to compare two PDF files, but to transform the information in the documents into verifiable data.

1. Identify the documents

The system begins by recognizing the documents in the folder:

• commercial invoice;
• packing list;
• Bill of Lading;
• order;
• other transportation or international trade documents.

This step ensures that each file is not treated as a standalone document. An international shipment consists of many items, received at different times and from different sources.

2. Extract the data

The documentary AI excerpt:

• references;
• descriptions;
• quantities;
• weight;
• package;
• values;
• dates;
• document numbers.

This is the traditional role ofOCR and, above all, of Intelligent Document Processing.

But extraction is just the starting point.

Information extracted automatically is not necessarily reliable in its context.

An invoice can very well show “12,500 kg” and a packing list “12,600 kg” without there having been a reading error.

The issue, therefore, is no longer what is written, but whether that information is consistent with the other information in the file.

3. Standardize the information

Before comparing the data, you need to be able to handle:

• the different units;
• formats;
• References;
• abbreviations;
• presentation conventions.

This step prevents the creation of artificial anomalies. For example, directly comparing a quantity expressed in metric tons with another expressed in kilograms could result in a false discrepancy. The automation process must first ensure that comparable information is being compared.

4. Align the lines

The same item may be listed differently on an invoice and a packing list. The system must identify which items actually correspond.

This is particularly important when documents contain multiple references or multiple levels of packaging.

An invoice may include:

Product A — 500 units

whereas the packing list lists:

Product A — boxes 1 through 10 — 500 units.

The structure is different, but the information refers to the same product.

5. Apply consistency checks

Once the data has been identified, standardized, and reconciled, the system can perform the checks.

For example:

Quantities

Invoice: 1,000,
Packing list: 1,000

→ Consistent

References

Invoice: SKU-4587
Packing list: SKU-4587

→ Consistent

Weight

Invoice: 12,500 kg
Packing list: 12,600 kg

→ Discrepancy detected

Package

Invoice: 50 cartons
Packing list: 52 cartons

→ Discrepancy to be analyzed

But automation shouldn't stop at this stage.

6. Identify discrepancies

A discrepancy can be:

• a real inconsistency;
• a missing data point;
• a normal difference;
• a problem with consistency;
• a situation that requires human verification.

The quality of automation therefore depends as much on the classification of anomalies as on their detection.

7. Trigger an action

Once the discrepancy has been identified, the system can:

• notify the operator;
• display the relevant documents;
• highlight the different values;
• propose an audit;
• recommend an action;
• uphold the decision that was made.

The goal is not to replace the expert, but to spare them the trouble of manually searching for anomalies throughout the entire file.

What types of discrepancies can be detected automatically?

A comparison of the invoice and the packing list can reveal several types of discrepancies.

Quantity Variance

Invoice: 1,000 units
Packing list: 960 units

→ Discrepancy to be analyzed.

The system searches for the quantity specified in the order or in other documents in the file.

Reference Deviation

Invoice: SKU-4587
Packing list: SKU-4589

→ The reference does not match.

This indicates a data entry error, an incorrect reference, or a change that has not been updated.

Designation Discrepancy

The two documents use different descriptions for the same commodity. The inspection must therefore be able to distinguish between a difference in description and an actual difference in the product.

Weight difference

Invoice: 12,500 kg
Packing list: 12,600 kg

→ Discrepancy detected.

But it is also important to verify the nature of the data:

• net weight;
• gross weight;
• declared weight;
• weight carried.

An effective check must therefore take into account the type of weight being compared.

Discrepancy in the number of packages

Invoice: 50 cartons
Packing list: 52 cartons

→ Discrepancy to be checked.

Once again, the context may explain the difference.

Missing data

Consistency isn't just about contradictory data.

A document may be incomplete:

• reference missing;
• quantity not specified;
• Expected document not received;
• Data required for a check that is unavailable.

Document verification must therefore ensure both consistency and completeness.

An invoice and a packing list do not necessarily have to "match"

This is probably the most important point for understanding the difference between a basic document review and a more advanced approach.

Let's talk about weight.

The invoice may include:

Net weight: 12,500 kg

The packing list:

Net weight: 12,500 kg
Gross weight: 12,600 kg

There is no inconsistency.

The two documents simply do not contain exactly the same information.

The same logic applies to quantities.

Invoice: 1,000 units
Packing list: 960 units
Order: 960 units

The system must be able to recognize that a discrepancy between the invoice and the packing list warrants verification, but that the order provides essential context.

That is why the goal is not to "match" two documents.

The goal is to verify that the information describing the same transaction is consistent.

‍

Why OCR and AI-powered data extraction aren't enough

It is helpful to distinguish between several levels.

OCR reads, IDP extracts.

But reconciliation makes it possible to cross-check information. This is an important distinction in document automation.

A solution can perfectly extract data from an invoice and a packing list without necessarily knowing whether that data is consistent between the two documents.

A piece of data may be extracted perfectly and yet be inconsistent with the rest of the file.

Why check for consistency at the shipment level?

An invoice and a packing list almost never go together. An international transaction also requires:

• an order;
• a shipping order;
• a bill of lading;
• a CMR or an AWB;
• a certificate of origin;
• customs documents;
• a shipping invoice;
• proof of delivery.

Each document contains a portion of the information. The problem arises when these different sources no longer match. That is why verification must no longer focus solely on:

Invoice ↔ Packing List

but gradually over:

Invoice ↔ Packing List ↔ B/L ↔ Order ↔ Operational Data

This is how we move from a document to a file.

The Logistics File: Preserving Context to Ensure Reliable Checks

A transportation case is evolving.

New documents are arriving. An invoice is corrected. A packing list is replaced. A validation is performed. An operational event may change the interpretation of a piece of data.

Treating each document separately means losing some of that context.

The file must contain:

• the source documents;
• the extracted data;
• the inspections conducted;
• the anomalies detected;
• validations;
• corrections;
• the history of decisions.

Auditing is no longer a one-time activity. It becomes an ongoing review of documentation.

What should you do when an inconsistency is detected?

Good automation shouldn't just display: "Mismatch"

It must make the anomaly understandable and actionable.

For example:

Inconsistent quantity

Invoice: 1,000 units

Packing list
,960 units

Order: 960 units

Analysis
Discrepancy between the invoice and the packing list. The order matches the packing list.

Recommended action
: Check whether the invoice needs to be corrected.

This gives the operator immediate access to the information needed to make a decision.

This approach is important in complex document-processing workflows: AI automates checks and focuses human intervention on exceptions, rather than seeking to systematically replace domain expertise.

Traceability is also essential: it must be possible to trace the source document, the relevant data, the detected anomaly, the control applied, and the decision made.

Toward Continuous Document Verification

The file is not static. A new document is added after the invoice and the packing list. A new version can replace an old one. Information is corrected. A validation can change the status of a check.

That is why document reconciliation should not be viewed as a one-time, ad hoc verification. It can become an ongoing review of the file.

From Comparison to Documentary Reconciliation

Ultimately, this is where the difference lies between simple document automation and a truly business-oriented approach.

Compare

Invoice: 1,000,
Packing list: 960,
s → Difference.

Reconcile

Invoice: 1,000,
Packing list: 960,
Order: 960,
→ Discrepancy identified and contextualized.

Improve reliability

Discrepancy analyzed → reference source identified → recommended action → decision documented.

This progression is at the heart of the Docloop approach. Making documents reliable is not just about extracting more data. It involves transforming scattered documents into verified, consistent information that can be directly used in business processes.

The real challenge isn't to match two documents

Automating the verification process between a commercial invoice and a packing list seems simple. But a truly reliable verification process must go much further:

Identify documents → extract data → standardize → match → verify → detect discrepancies → contextualize them → recommend an action.

The difference is significant. A comparison seeks out a difference.

Reconciliation seeks to understand this difference.

And document reliability goes even further: it ensures that the information used to manage a logistics operation is consistent, complete, and actionable.

That is why true document automation is no longer just about transferring information from a PDF to a TMS or an ERP system.

It involves verifying that this information is reliable before passing it on in the process.