Home · Solutions · Supply chain
Solution · Supply chainThe customs pack is built from the delivery, not retyped into the broker's portal
Export documents ready while the truck is loaded
Robots build the commercial invoice, packing list and declaration data from your own order and delivery records, apply the code and origin rules, and file every confirmation.
Executive summary
Your ERP holds every figure the customs pack needs. Somebody types them in again anyway.
The flow starts where the data already lives.
The document set is ready when picking finishes instead of after it, so paperwork stops being the reason a truck waits on the yard.
the channel your customs agent or declaration software accepts; SAP S/4HANA (reference and status on the delivery); SharePoint shipment file
Business problem
Trade compliance
Every export shipment needs a document set before it can leave: a commercial invoice, a packing list that matches the pallets, certificates where the product or the destination requires them, and a declaration data set with commodity codes, origin, procedure and consignee. Almost all of it exists already, in the sales order, the delivery and the warehouse system, but none of it moves by itself, so a person opens both screens and types the same figures into a sheet, a template and a portal.
Classification and origin are the parts that hurt. A commodity code is a judgement made once and reused for years, and new material numbers arrive faster than anyone maintains the list, so the working rule becomes "take the code from the nearest similar part". Preferential origin depends on supplier declarations with validity dates, kept in a mail archive rather than against the material. Restricted goods depend on a product characteristic the clerk cannot see on the delivery note.
What makes this expensive is the feedback loop. Nothing tells the desk it was wrong: the pack is accepted, the truck leaves, and the error surfaces weeks later through the agent, the customer's broker or an authority. By then the goods have arrived and the invoice is paid, so the correction is an amended declaration and a conversation with a customer being charged duty on a machine bought last spring.
How it works today
Whatever the ERP and whoever the agent, this is the route we find on desks that grew with the business.
- PersonThe shipping clerk copies consignee, values, line data and weights from the outbound delivery into the sheet the agent accepts
- PersonCommodity codes come off a spreadsheet on a shared drive; a material not on it gets the code of the nearest recognisable part
- WaitingWhere a preference is claimed, the supplier's declaration of origin is hunted for in mailboxes and folders, and the pack waits until it turns up
- PersonInvoice and packing list come from Word and Excel templates, printed for signature and scanned back
- SystemThe sheet goes to the agent by email, or the figures are typed again into the agent's portal
- Risk of errorWrong codes, restricted items and expired authorisations are caught only if the agent notices, usually weeks later
- PersonThe reference and the confirmation that the goods left arrive by email and are filed when there is time
Why the current process costs more than it appears
The most expensive part of this process has no cost line.
- Retyping is the cheapest part of this process and the only part anyone counts; the searching around it, for a code, a certificate or a weight, is unrecorded and larger.
- Classification errors are discovered on somebody else's schedule, and the timing costs more than the correction: amended declarations, duty the customer did not expect, an account manager explaining paperwork to a buyer who cares about machines.
- Preference claims made without valid supplier evidence are a liability with a long fuse: the customer takes the benefit at import, the exporter carries the statement, and the evidence has to exist years later on request.
- Missing exit confirmations turn a shipping problem into a finance problem, because zero rating an export supply rests on showing the goods left, and a confirmation unread in a mailbox is not a filed document.
Cost of inaction
Row three is one scenario with its arithmetic on show, not a probability. Take 240 shipments of one product line in a year at an average invoice value of €15,000: that is €3.6m of goods, and three percentage points of duty on it is €108,000 argued about later, usually by the customer whose broker found it.
What compounds is not the salary line. Each month the classification file gains rows nobody reviewed, the reasoning behind older ones fades, and the two people who can reconstruct it get harder to replace. Such a desk does not fail visibly; it simply cannot say how many of last year's codes would survive a check.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A European machinery manufacturer exporting to 40 countries from two plants and a spare-parts warehouse, on SAP S/4HANA and Microsoft 365 E3 with a separate warehouse system; one four-person export desk serves both plants.
1,300 export shipments a month: complete machines, project shipments with many lines, and a long tail of spare parts; roughly 300 new or changed material numbers each quarter.
Data is copied from the delivery into the agent's sheet, codes come from a spreadsheet one person maintains, the pack is made from templates, and confirmations are filed in a quiet hour.
About 11 minutes of document work per shipment before anything is chased, a backlog whenever a plant ships a project order with 60 lines, and a code list a few hundred materials behind.
Robots build the shipment data set from the delivery and the warehouse system, apply the classification, origin and restriction rules, generate the pack, hand the declaration data to the agent, and file what comes back.
In the modelled case the standard flow is prepared without a person, the desk sees only classification and restriction exceptions, and each shipment carries its evidence from the day it left. The figures are a model, not a measurement.
Proposed solution
The flow starts where the data already lives. A released outbound delivery creates a queue item, and a robot collects what a document set needs: consignee and destination, Incoterms, line items with values and quantities, and from the warehouse system the actual weights and handling units. That set is tested against a classification table your trade compliance officer owns, carrying commodity code, origin, procedure, preference status and restriction flag per material. Four things stop a shipment: a material with no entry, a preference claimed without evidence, a restriction flag, a destination needing an authorisation.
Supplier evidence arrives as documents, so it is read as documents. Certificates of origin, supplier declarations and contract-manufacturer packing lists are extracted by UiPath Document Understanding, which ships with pre-trained models for these types, and the fields are written against the material with their validity dates. Low-confidence readings go to Validation Station rather than into the pack unverified. Sanctions screening of the consignee is a control in its own right, so we connect to it rather than rebuild it here.
The pack is then generated from your own templates and the declaration data set handed over. We are deliberately narrow here about what we promise: the data goes to the channel your customs agent or declaration software accepts, a portal upload, a structured file or their own API, and we do not connect to a national customs system on your behalf. What comes back, the reference and later the confirmation that the goods left, is written to the delivery and filed with the pack, while whatever a rule stopped becomes a task in Microsoft Teams.
UiPath Document Understanding pre-trained models for Certificate of Origin, Packing Lists and EU Declaration of Conformity, with Validation Station; UiPath Orchestrator queues, triggers, credential stores and audit; UiPath Integration Service connectors for SAP, Microsoft Teams and Microsoft OneDrive & SharePoint; UiPath Action Center tasks in Microsoft Teams; Microsoft Purview retention labels
The shipment data set out of the ERP and warehouse system, the classification, origin, procedure and restriction rules, document generation from your templates, the agent hand-off, the write-back of references and confirmations, the exception flow, the shipment file
Your customs agent's or declaration provider's interface, through UiPath Integration Service Connector Builder where an API exists and as a file or portal exchange where it does not; warehouse weights and handling units where no connector exists
How the automated process works
- AutomationA released outbound delivery creates a queue item; the robot collects order, consignee and value data from the ERP and weights and handling units from the warehouse system
- SystemThe classification table supplies commodity code, origin, procedure and preference status per material, and flags anything it does not know
- AutomationDocument Understanding reads the supplier certificates and declarations behind those materials, and the rules run: restriction flags, authorisation validity, preference evidence, destination thresholds, completeness of the line data
- PersonAnything a rule stops reaches the trade compliance officer as an Action Center task in Microsoft Teams, with the material, the reason and a proposed answer
- AutomationInvoice and packing list are generated from your templates, and the declaration data goes to the channel your agent or declaration software accepts
- AutomationThe reference and the exit confirmation are written back to the delivery and filed with the pack in the SharePoint shipment file, and a daily summary reaches the desk's Teams channel
Human-in-the-loop model
Automation handles
- Building the shipment data set from the ERP and the warehouse system, weights and handling units included
- Applying the classification, origin, procedure and restriction rules to every line of every shipment
- Generating the pack, handing the declaration data to the agent, and filing the reference, confirmation and rule trail
People decide
- The commodity code for a material the table does not know, and any code the rules propose changing
- Whether a shipment carrying a restriction flag or an expiring authorisation may leave
- Whether a preference is claimed when the supplier's declaration is missing, expired or unreadable
- The rules themselves: table, thresholds, destination groups and templates stay owned by trade compliance
Before and after
Systems and integrations
Everything below runs on licences and systems you already hold, or would need anyway.
Inputs
- ERP sales orders and outbound deliveries
- warehouse weights, packaging and handling units
- supplier certificates and declarations arriving by email
- your classification and origin table
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Document Understanding
- UiPath Integration Service
- UiPath Action Center
Target systems
- the channel your customs agent or declaration software accepts
- SAP S/4HANA (reference and status on the delivery)
- SharePoint shipment file
Human touchpoints: Action Center tasks in Microsoft Teams; Validation Station for unreadable documents; the export desk's daily Teams summary
Technologies used
one queue item per shipment; rules, retries, credentials, audit trail
Apre-trained models read certificates of origin, conformity declarations and supplier packing lists; Validation Station for low-confidence fields
Areads orders and deliveries, writes references back, posts tasks, files the pack
Aclassification and restriction exceptions decided where the officer works
Athe shipment file: pack, rule trail, reference and confirmation, retained
Athe 8-digit code list your table is kept against, republished annually
Bthe reference behind the restricted-goods check
BIllustrative economic model
A model, not a promise.
What the 11 minutes cover is the paperwork alone: pulling the data together, filling the agent's sheet, generating the pack, filing what comes back. Packing, loading, the agent's own work and the unmeasured time spent chasing a code sit outside the model. It is a blended average, since a spare-part shipment takes far less and a project order far more. €30 is an illustrative fully loaded hourly cost for a shipping role in Central Europe, and nothing here was measured at a client.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- The document set is ready when picking finishes instead of after it, so paperwork stops being the reason a truck waits on the yard
- Codes, origin, procedure and restriction status come from one maintained table applied identically to every shipment, whoever is on shift
- Restricted goods, missing authorisations and preference claims without evidence stop the pack rather than the truck, before the goods move
- Every shipment leaves with its evidence complete, so last year's delivery is answered from a file rather than a search
- The desk's experienced people spend the day on classification and licences, not on copying figures
The management view
- Shipment paperwork becomes a measured process: what was prepared automatically, what stopped, on which rule, and how long the decision took
- Volume peaks and new destination countries are absorbed by configuration and queue capacity, not by overtime and a new routine to learn
- Knowledge that lives in two people's judgement becomes a table with an owner, a history and a reviewer, and every shipment either has its confirmations filed or appears on a list
Board-level KPIs
Security and governance
Where the data sits and who can see it.
- Robots read the ERP and the warehouse system with dedicated accounts holding display rights plus the single write the reference needs, and the agent's portal is reached under a service identity whose secret sits in the credential store, not in a workflow
- The shipment file keeps the pack, the rule trail, the reference and the exit confirmation together under a Purview retention label, so evidence outlives whoever created it
- Classification and restriction rules are versioned and compliance-approved, and each run records which version decided the shipment
- Orchestrator logs every robot action; the documents, tasks and rules it touches stay in your Microsoft 365 tenant and UiPath Automation Cloud, EU region
Why now
The tariff has a published expiry date. Annex I to Council Regulation (EEC) No 2658/87, the Combined Nomenclature with its 8-digit codes, is updated and republished every year in the EU's Official Journal, so a hand-kept code list goes stale on a schedule and nobody gets a reminder.
Restricted goods are an operational question now, not only a legal one. Regulation (EU) 2021/821 governs dual-use exports, its Annex I control list was last updated by delegated regulation in September 2025, and the record keeping it expects is what a manual desk cannot produce. As salaried work, the document handling is worth about €7,150 a month.
The technical parts stopped being projects: pre-trained models for certificates and packing lists, standard connectors into the ERP and Microsoft 365, exceptions decided in a Teams window already open.
Relevant executive roles
Shipments stop waiting on paperwork, and a project order with sixty lines stops being the reason a plant misses a truck slot
Evidence behind zero-rated export supplies is filed with the invoice instead of sitting in a mailbox, and duty corrections arriving months later as customer claims become rare
Classification becomes a maintained table with a history rather than a spreadsheet nobody dares touch, and every decision on a stopped shipment is attributable
Common questions and objections
It does not. The automation applies the table your compliance officer owns, and where a material has no entry it proposes the nearest match with its reasoning and stops the pack. A person confirms the code, which then enters the table.
Where the data comes from, and what happens before it leaves you. Today the agent gets a sheet typed by a clerk and their checks are the only control. Afterwards they get a set built from the delivery and tested against your rules.
Normal, and the reason to start with one. What differs, meaning templates, agents, destinations and procedures, sits in configuration rather than code, so the second site is a rollout.
When this is not the right solution
- Fewer than roughly a hundred export shipments a month, where a disciplined routine and good templates cost less than building rules
- A material master without usable classification data, where the first project is putting codes and origin onto the material record rather than generating documents from it
- Declarations lodged by your own staff in a national system with no interface for third-party software; the pack and the checks still pay, but the submission stays manual
A question for the next management meeting
When a customer's broker questions a commodity code we used last year, how long does it take us to produce the shipment, the evidence behind the code and the person who chose it?
Implementation approach
The first week looks the same at every client: we look at the data.
We deliver
- The rules layer: classification table, origin and preference logic, procedure and restriction checks, owned by your compliance officer
- Document generation from your templates and the hand-off to your agent's portal, file or API
- Exceptions in Action Center and Microsoft Teams, the SharePoint shipment file with retention, the daily desk summary
- A pilot on one destination group, then rollout per site with hypercare and a runbook for the desk that uses it
We need from you
- Three months of export shipments with the document sets that went with them
- Your classification table as it stands, and the supplier origin declarations you rely on
- A trade compliance owner who can decide codes, thresholds and what a restriction flag must do
- Technical accounts for the ERP, the warehouse system and your agent's channel
Stages
Discovery
One month of shipments read end to end: documents, destinations, procedures, exceptions, volumes
Rules
Classification, origin, procedure and restriction logic written with trade compliance, tested on past shipments
Build
Data collection, document generation, the agent hand-off, Teams exceptions, the shipment file
Parallel run
Packs generated beside the manual ones and compared line by line before anything is sent
Go-live
One destination group first, then site by site, with hypercare and a weekly rules review
Departmental. Effort follows the number of destination groups and procedures, the state of the material master, and how your agent accepts data.
The truck leaves Thursday. The commodity code was chosen on Wednesday evening.
Send us one month of export shipments: destinations, document sets, and where the codes and weights come from today. You get back a count of the materials with no reliable code, the exceptions your desk would keep, and a first estimate.
Find the materials without a codeThe neighbouring process usually has the same problem
Rework, scrap and freight caused by a supplier part are absorbed internally because the claim costs too much to build.
View solution Supply chainCustoms declarations that clear on the first submissionA truck waits while a broker refreshes a screen. Most of those rejections were visible in the documents.
View solution Legal & complianceCounterparty screening: sanctions, VAT status, registersYour counterparty file was true on the day you checked it. Sanctions and VAT status moved after that.
View solution Supply chainCustomer orders from PDFs and emails straight into the ERPCustomer POs still get retyped from PDFs into SAP and confirmed hours later. Stop paying your order desk to type.
View solution Case studyDigital CMR and PODThe driver photographs the document at unloading.
View case study Case studyDriver settlementNo more manual document collection and spreadsheets.
View case studyIndustries we deliver this in most oftenManufacturing & industryTransport & logisticsRetail & e‑commerce