Customs Explorer
⌕

Customs Explorer

Document automation

CUSTOMS AUTOMATION / DOCUMENT INTELLIGENCE

Document
Automation.

Turning customs documents into structured, searchable and actionable information.

Customs operations depend on commercial invoices, packing lists, transport documents, certificates, authorisations and supporting records. Document automation helps organisations capture, organise and validate this information while reducing repetitive manual handling.

The objective is not simply to digitise documents. It is to connect documentary evidence with the underlying customs transaction, identify discrepancies and make relevant information available to authorised operators and systems.

01 Document capture Collect and organise source records
02 Data extraction Identify relevant documentary information
03 Controlled review Resolve exceptions before operational use
DOCUMENT PROCESSING From files to customs-ready data
CONCEPTUAL FLOW
DOCUMENT INPUT Commercial & supporting records Invoices · Packing lists · Transport documents
DOCUMENT INTELLIGENCE Classification & information extraction Document type · Key fields · References
QUALITY CONTROL Validation & discrepancy detection Missing fields · Conflicts · Human review
CONTROLLED OUTPUT Structured document information Linked records · Review status · System handoff

The central challenge: transforming fragmented documentary information into reliable data without losing the original evidence, its context or the responsibility for customs decisions.

01 / DOCUMENT PROCESSING WORKFLOW

From documents to
usable customs information.

Document automation follows a controlled sequence: acquiring source records, recognising their content, extracting relevant information, validating results and making approved data available to customs operations.

PROCESS ARCHITECTURE

Five connected stages.

Each stage should preserve the relationship between the original document, the extracted information and the customs transaction to which it belongs.

Core principle Automating data extraction does not remove the need to verify the accuracy and relevance of the underlying information.

01
DOCUMENT INTAKE

Acquire and register documents

Collect records from email attachments, document repositories, ERP systems, supplier portals or other approved sources. Register each file with a unique reference and available metadata.

EXAMPLE INPUTS
PDF invoices Packing lists Transport records
02
DOCUMENT RECOGNITION

Identify document type and structure

Determine whether a record is a commercial invoice, packing list, transport document, certificate or another supporting record. Different document layouts may require different extraction methods.

PROCESSING METHODS
Text parsing OCR where needed Layout recognition
03
INFORMATION EXTRACTION

Convert document content into structured fields

Extract relevant information such as invoice references, parties, dates, product descriptions, quantities, values, currencies and shipment identifiers. Link extracted fields to their source documents and, where feasible, their original locations.

STRUCTURED OUTPUT
Invoice reference Extracted field
Currency Extracted field
Goods quantity Extracted field
04
QUALITY ASSURANCE

Validate results and resolve discrepancies

Apply configured checks for missing information, inconsistent values, unexpected formats and mismatches between related records. Route uncertain extraction results and material discrepancies to a reviewer.

EXAMPLE CHECKS
Missing fields Invoice totals Quantity mismatches
05
CONTROLLED HANDOFF

Connect approved information to customs operations

Make reviewed information available to authorised users, declaration preparation workflows or compatible business systems. Retain links to source documents, validation results and relevant review decisions.

DOWNSTREAM USE
Declaration preparation Document review Record retrieval
OPERATIONAL VALUE

What changes when documents become structured data?

01 Less repetitive entry

Information can be reused across approved workflows instead of being retyped from each source document.

02 Earlier discrepancy detection

Missing or inconsistent information can be identified before it disrupts subsequent customs processing.

03 Better documentary traceability

Extracted fields remain associated with their source records, making verification and retrieval easier.

Important distinction: document automation prepares and checks information. It does not independently prove that a document is authentic, that an origin claim is valid or that the evidence satisfies every applicable customs requirement.

02 / DOCUMENT TYPES & EXTRACTION

Different documents.
Different information.

Customs documentation is not a uniform dataset. Commercial, logistical and regulatory records serve different purposes, contain different fields and require different levels of interpretation. Effective automation must recognise these distinctions.

DOCUMENT CATEGORIES

Six documentary sources commonly encountered in customs operations.

The examples below illustrate potential extraction targets. They do not imply that every listed document is mandatory for every customs procedure.

01 / COMMERCIAL

Commercial invoices

Commercial transaction records often provide important information for declaration preparation and customs valuation analysis.

Potential extraction fields
  • Seller and buyer information
  • Invoice number and date
  • Goods descriptions and quantities
  • Prices, totals and currency
  • Stated delivery terms

Invoice values do not necessarily equal customs value.

02 / PACKAGING

Packing lists

Packing information supports the reconciliation of commercial goods, packages and shipment quantities.

Potential extraction fields
  • Package references and counts
  • Goods or item identifiers
  • Quantities per package
  • Gross and net weights
  • Package marks and dimensions

Package data should be reconciled with the shipment and invoice.

03 / TRANSPORT

Transport documents

Transport records help identify consignments, carriers and logistical references relevant to the customs process.

Potential extraction fields
  • Transport document reference
  • Consignor and consignee
  • Transport mode and routing
  • Place of loading or delivery
  • Shipment and package details

Available fields depend on the transport document type.

04 / ORIGIN

Origin-related evidence

Origin statements, certificates and supplier declarations may provide information relevant to origin assessments and preferential claims.

Potential extraction fields
  • Document or statement reference
  • Issuer or exporter information
  • Goods covered by the evidence
  • Origin declaration wording
  • Dates and relevant identifiers

Extracted origin wording does not establish eligibility for preference.

05 / REGULATORY

Licences & certificates

Certain goods and procedures may require licences, certificates or other regulatory supporting documents.

Potential extraction fields
  • Document type and reference
  • Issuing authority
  • Holder or beneficiary
  • Validity dates
  • Goods or quantities covered

Authenticity, validity and applicability require separate checks.

06 / CUSTOMS RECORDS

Customs references & decisions

Existing customs records, decisions and authorisations can supply references for subsequent operations and documentary checks.

Potential extraction fields
  • MRN or other reference numbers
  • Decision or authorisation references
  • Holder and issuing authority
  • Procedure-related information
  • Dates and recorded statuses

A reference found in a document does not confirm its current status.

TECHNOLOGY & INTERPRETATION

Reading a document is not the same as understanding it.

Document processing involves distinct technical and professional activities. Combining them without clear boundaries can create misleading confidence in automated results.

LAYER 01

Text recognition

OCR & TEXT EXTRACTION

Convert scanned content into machine-readable text, or extract text directly from digital documents when supported.

EXAMPLE Recognise an invoice number printed on a scanned PDF.
Does not establish

Whether the recognised characters are correct or relevant.

LAYER 02

Structured extraction

FIELD IDENTIFICATION & MAPPING

Identify the meaning of recognised content, associate values with fields and transform them into a defined data structure.

EXAMPLE Associate a monetary amount with the invoice total and currency.
Does not establish

Whether the extracted value is legally sufficient for customs use.

LAYER 03

Customs assessment

RULES, EVIDENCE & HUMAN JUDGEMENT

Assess the extracted information against applicable customs requirements, available evidence and the circumstances of the operation.

EXAMPLE Determine whether an origin statement supports a particular claim.
Requires

Appropriate legal criteria, evidence and accountable decisions.

PRACTICAL EXAMPLE

One invoice. Three different questions.

01 / RECOGNITION What does the invoice say?

The software reads a goods description, quantity and invoice amount.

02 / EXTRACTION Which data fields do these values represent?

The system maps the recognised values to product, quantity and commercial amount fields.

03 / ASSESSMENT Are those values sufficient for the customs operation?

The operator evaluates classification, valuation and other relevant requirements using the available evidence.

Implementation principle: reliable document automation should retain source references, distinguish extracted facts from inferred information, flag uncertainty and support correction without losing the original documentary evidence.

03 / PRACTICAL APPLICATIONS

From document processing
to operational value.

The value of document automation becomes visible when extracted information is compared, verified and connected to a specific customs operation. The following examples illustrate controlled workflows rather than automatic compliance decisions.

SCENARIO 01 / COMMERCIAL DOCUMENTS 01

Invoice and packing list reconciliation.

A shipment arrives with a commercial invoice and a packing list. Both contain overlapping information, but their structure and level of detail may differ.

COMMERCIAL INVOICE
Item reference Extracted
Quantity Extracted
Commercial value Extracted
PACKING LIST
Item reference Extracted
Quantity Extracted
Package / weight Extracted
DOCUMENT EXTRACTION FIELD MATCHING EXCEPTION REVIEW
What automation can detect

Differences in item references, quantities or package information, provided that the relevant fields can be reliably extracted and matched.

Potential benefit: discrepancies can be investigated before information is transferred into a declaration workflow.

SCENARIO 02 / ORIGIN DOCUMENTATION 02

Origin evidence identification and review.

A shipment includes origin-related documentation. The system identifies the record and extracts available references, statements and product information for further assessment.

01
Identify the evidence

Recognise the document type and capture the relevant statement or certificate information.

02
Associate the goods

Match documentary references to the products or shipment concerned, where possible.

03
Check documentary consistency

Flag missing references, unclear wording, conflicting identifiers or other configured issues.

04
Route for customs assessment

Make the evidence available for review against the applicable origin rules and the circumstances of the transaction.

What automation cannot assume

The presence of an origin statement or certificate does not independently establish preferential origin, document validity or eligibility for preferential treatment.

Potential benefit: relevant origin evidence becomes easier to locate, associate with goods and review.

SYSTEM INTEGRATION

Connecting documents to enterprise workflows.

Document automation can operate as an intermediate processing layer between source repositories, business applications and customs workflows. The architecture must be adapted to actual system interfaces and access permissions.

SOURCE SYSTEMS

Document repositories

Email attachments
ERP document records
Approved storage systems

PROCESSING LAYER

Document automation

Recognition & extraction
Field mapping & validation
Exception routing

CONTROLLED PROCESSING
DOWNSTREAM SYSTEMS

Customs operations

Declaration preparation
Documentary review
Audit and record retrieval

01 / CONNECTIVITY

Available interfaces

Verify whether source and destination systems provide supported APIs, file exchanges or other integration mechanisms.

02 / DATA MODEL

Field definitions

Establish how document fields, identifiers and relationships map to the target operational data structure.

03 / SECURITY

Access and confidentiality

Define permissions, authentication, appropriate data handling and protection of commercially sensitive records.

04 / OPERATIONS

Maintenance and exceptions

Account for changing document layouts, extraction failures, integration monitoring and ongoing processing costs.

IMPLEMENTATION PRIORITY

Start with a narrow, measurable workflow.

01 Select a document set

Choose a repeatable transaction type with sufficiently consistent source records.

02 Define the required fields

Specify the data to extract and the checks needed before operational use.

03 Measure actual performance

Evaluate extraction accuracy, exception rates, review time and the cost of corrections.

Technical feasibility comes first: before committing to a document automation platform, verify integration access, document formats, data protection requirements, processing costs and the limits of the proposed extraction technology.

04 / CONTROLS & EVIDENCE INTEGRITY

Automate document handling.
Preserve the evidence.

A document automation system should not be evaluated only by how quickly it extracts information. It must also protect source records, identify uncertainty, manage corrections and preserve a reliable connection between documentary evidence and customs decisions.

OPERATIONAL RISK MATRIX

Six controls that matter.

Each control addresses a distinct failure mode. The appropriate implementation depends on document sensitivity, transaction volume, regulatory requirements and the systems involved.

01 / EXTRACTION ACCURACY

Incorrectly extracted information

OCR and extraction systems may misread characters, confuse fields or associate values with the wrong document section.

CONTROL RESPONSE Field validation, uncertainty flags and targeted review.
02 / DOCUMENT IDENTITY

Duplicate or incorrectly matched documents

The same record may be received more than once, or a document may be linked to the wrong shipment or commercial transaction.

CONTROL RESPONSE Document identifiers, duplicate detection and transaction matching.
03 / VERSION MANAGEMENT

Outdated or superseded records

Revised invoices, corrected packing lists and replacement certificates may coexist with earlier versions.

CONTROL RESPONSE Version history, supersession status and controlled document selection.
04 / EVIDENCE TRACEABILITY

Loss of the original evidence

Extracted values may become detached from the source files needed to verify them or explain subsequent corrections.

CONTROL RESPONSE Persistent source links, controlled storage and provenance records.
05 / ACCESS SECURITY

Unauthorised access or modification

Commercial invoices and customs documents can contain sensitive business information and records subject to access restrictions.

CONTROL RESPONSE Role-based permissions, protected storage and access logging.
06 / SYSTEM RELIABILITY

Silent processing or integration failures

A file may fail to process, an extraction may remain incomplete or a downstream system may reject the transferred data.

CONTROL RESPONSE Processing states, error monitoring and controlled retry procedures.
HUMAN-IN-THE-LOOP CONTROLS

A review process that remains accountable.

Automated extraction can support review, but material uncertainty and customs-relevant discrepancies should be directed to people with the appropriate responsibilities and knowledge.

INPUT

Processed document

Original file, extracted fields and processing metadata.

CHECKPOINT 01

Extraction verification

Review uncertain fields, missing information and extraction errors.

CHECKPOINT 02

Documentary assessment

Assess relevant inconsistencies, evidence and required approvals.

CONTROLLED OUTPUT

Approved information

Reviewed data and source references available for authorised use.

Approval has a defined scope. Approval of extracted data confirms only the checks actually performed. It should not be represented as confirmation of document authenticity, preferential origin, customs valuation or other legal conclusions unless those matters were specifically assessed.

PROVENANCE & AUDIT TRAIL

Every extracted value needs a traceable origin.

A defensible document workflow should make it possible to understand which source was processed, what the system extracted, what changed and who approved the resulting information.

01

Source document

Record the original file, document reference, receipt information and its association with the relevant transaction.

02

Processing history

Record extraction runs, processing outcomes, relevant software or rule versions and reported exceptions.

03

Corrections & approvals

Preserve relevant changes, their reasons, responsible users and the status of required reviews.

04

Downstream use

Record which approved dataset or document version was made available to another system or operational process.

EVIDENCE INTEGRITY

Protect the relationship between the original and the processed record.

Document identifiers, file hashes, access logs and version records can support integrity and traceability when appropriately implemented. A hash can help detect changes to file content, but it does not establish that the original document was authentic or legally valid.

Retention periods, confidentiality requirements and permitted access must be assessed according to the applicable customs, contractual and data protection obligations.

The objective is controlled information, not blind automation. A reliable system makes documents easier to process while keeping errors visible, decisions attributable and supporting evidence available for appropriate review.

05 / QUESTIONS, REGULATION & NEXT STEPS

Document automation.
Questions that matter.

Understanding what document automation can do is only part of the picture. Organisations must also consider legal responsibilities, evidence requirements, system integration and the limits of automated processing.

FREQUENTLY ASKED QUESTIONS

Six essential questions.

Practical answers for businesses evaluating document automation within EU customs operations.

01 / AUTOMATION SCOPE

Can customs documents be processed automatically?

Yes. Software can collect files, recognise document types, extract information, compare records and route exceptions. The achievable level of automation depends on document quality, available data, integration capabilities and the controls required.

02 / OCR & AI

Is OCR the same as document automation?

No. OCR converts image-based text into machine-readable characters. Document automation can also include classification, field mapping, validation, workflow routing, approval and controlled distribution. OCR is one possible component of that process.

03 / CUSTOMS COMPLIANCE

Does automated extraction make a document legally valid?

No. Extracting information does not establish authenticity, validity or legal sufficiency. Customs-relevant documents must still be assessed against the requirements applicable to the goods, procedure and transaction concerned.

04 / DECLARATION PROCESSING

Can extracted document data populate a customs declaration?

It can support declaration preparation where suitable interfaces and mappings exist. However, data must be checked against the applicable declaration requirements. Preparing information is distinct from submitting a declaration through an authorised customs system.

05 / DOCUMENT RETENTION

Can original documents be deleted after extraction?

Not simply because their data has been extracted. The applicable recordkeeping obligations, evidential requirements and retention rules must be considered. Article 51 of the Union Customs Code addresses the keeping of documents and information for customs control purposes.

06 / IMPLEMENTATION

What should a business verify before adopting document automation?

Start with actual document samples and required data fields. Verify extraction accuracy, supported interfaces, information security, review procedures, exception rates, operating costs and the ability to preserve source-document traceability.

EU REGULATORY FRAMEWORK

The legal framework behind customs information.

Document automation is a technical process, not a separate customs authorisation. Its outputs must be handled consistently with the legal requirements governing customs declarations, supporting documents and recordkeeping.

PRIMARY LEGISLATION 01

Union Customs Code

Regulation (EU) No 952/2013

Establishes the core customs framework. Relevant provisions include Article 15 on information supplied to customs authorities, Article 51 on recordkeeping and Article 163 on supporting documents.

Read on EUR-Lex
DELEGATED RULES 02

UCC Delegated Regulation

Commission Delegated Regulation (EU) 2015/2446

Supplements the Union Customs Code with detailed rules, including provisions relevant to customs procedures and data requirements. The applicable provisions depend on the operation concerned.

Read on EUR-Lex
IMPLEMENTING RULES 03

UCC Implementing Regulation

Commission Implementing Regulation (EU) 2015/2447

Sets out procedural and technical implementing rules for the Union Customs Code. It provides part of the regulatory context for customs data and related operational processes.

Read on EUR-Lex
THE BIGGER PICTURE

Better document processing starts with better controls.

Automation is most useful when source evidence remains accessible, extracted information is verifiable and customs-relevant decisions stay within a defined accountability framework.

Explore CustomsTech & AI