Automated customs declarations for UK-EU freight forwarders
Since the UK/EU Safety & Security (ENS) rules came in, every truck crossing the border carries 45 minutes of manual data entry - and a single mis-keyed HS code or weight discrepancy can cost £1,000s in fines and held shipments. Zeroflowers builds an end-to-end document workflow - ingestion, extraction, validation, multilingual generation, audit - that takes invoices, packing lists and bills of lading from inbox to compliant declaration without a human re-typing anything.
Where freight forwarders lose hours and money
- Operators re-key data from commercial invoices and packing lists into the customs portal - 45 minutes per truck.
- HS codes and weights are mis-typed under time pressure - the truck is held at the border.
- Mobile photos and blurry scans of paperwork break standard AI extractors.
- Invoice, packing list and bill of lading aren't cross-checked - discrepancies surface at the border, not in the office.
- There's no audit trail showing where each declared value came from when HMRC asks.
What an end-to-end customs workflow looks like
- Inbox monitor that picks up incoming shipment docs and matches them to a job in your CRM or TMS (e.g. CargoWise).
- Document assistance agents (our 'messy extractor') that pull HS codes, net/gross weights and values from blurry mobile photos with cited bounding boxes.
- Automated 3-way match between Invoice, Packing List and Bill of Lading - discrepancies flagged for human review before the truck moves.
- Paperless generates the final compliant declaration in EN + LT (or your destination languages) in milliseconds.
- Immutable governance dashboard showing exactly which pixel of which document each value came from - audit-ready for HMRC inspections.
How we build the customs declaration assembly line
This is the same 5-stage pattern we ship for every compliance-led document workflow - here applied to UK/EU customs. AI agents are stage 2 (extraction). They are not the product; the workflow is.
1. Ingestion
An email/inbox monitoring agent picks up incoming shipment docs (PDFs, scans, mobile photos) and matches each one to a specific job reference in CargoWise, your CRM or your TMS.
2. Extraction
Our document assistance agent (the 'messy extractor') pulls structured data - HS codes, net/gross weights, declared values - from blurry scans or tilted mobile photos. Every value comes with a page + pixel citation.
3. Validation
A 3-way match cross-references Invoice total vs Packing List vs Bill of Lading. If invoice weight = 500 kg but packing list = 505 kg, the system flags it for human review immediately - not at the border.
4. Generation
Once validated, Paperless generates the final compliant customs declaration in English and Lithuanian (or your destination languages) in milliseconds - one source of truth, every required output.
5. Audit & Governance
An immutable dashboard shows where every data point was extracted from - the exact pixel of the exact PDF. HMRC and ENS inspectors get a clean trail; you get peace of mind.
Document workflow vs general AI chat
A general-purpose AI chat tool can read one document well. A Zeroflowers customs workflow runs 24/7 across thousands of documents per month with cited extractions, integrated logic and an audit trail HMRC will actually accept.
| Feature | General AI (ChatGPT / Claude / Gemini) | Zeroflowers document workflow |
|---|---|---|
| Trust & hallucination | May 'guess' a number if the scan is blurry; no source citation. | Visual grounding - only extracts a value if it can prove the page + pixel coordinates it came from. |
| Workflow vs chat | Requires a human to copy-paste each document into a chat window and re-type the answer back into the customs portal. | End-to-end API that runs 24/7 - inbox -> extraction -> match -> generation -> audit, no human in the loop on the happy path. |
| Complex business logic | Can't easily check a PDF against a live database or ERP. | Cross-references extracted data against your CargoWise, TMS or SQL database in real time. 3-way matches built in. |
| Multilingual context | Generic translations that risk mis-stating regulated terms. | Generates legally accurate technical declarations in EN + LT (or your locales) from one source of truth. |
| Auditability | No record of why a value was suggested. | Immutable per-field audit trail - regulators see exactly where each value came from. |
| Data ownership | Public model providers may train on your sensitive trade data unless you self-manage policies. | Proprietary infrastructure - logic and data stay in a controlled EU environment with DPAs and zero retention. |
This is why a UK-EU freight forwarder pays £15-25k for the build plus a monthly retainer: it protects them from £100k in potential fines and held trucks - something a £20/month ChatGPT subscription cannot do.
Every document automation we ship runs on Paperless - our in-house engine, also available as a standalone product. Generate 10,000+ branded PDFs per month from CSV, Google Sheets or your REST API. Free to start.
- Visual no-code template editor your team can edit without a developer
- REST API for on-demand generation from any system
- CSV / Google Sheets batch generation - 10,000+ docs per run
- Multilingual output (EN / LT / PL / +) from a single template
- Free tier - no credit card required
Common questions
Will this satisfy HMRC and ENS audit requirements?
+
Yes - that's the point of stage 5 (Audit & Governance). Every declared value carries a citation back to the page and pixel of the source document. HMRC inspectors get a clean per-shipment trail; your operations team gets a single dashboard.
Can it integrate with CargoWise (or our TMS)?
+
Yes - CargoWise is one of the most common integrations we ship for freight clients. We also integrate with most TMS, customs broker portals and ERP systems via REST APIs and webhooks.
What happens when an extraction is low confidence?
+
The field is flagged for human review rather than guessed. Operators see exactly which page and area of the document was uncertain and can confirm the value in seconds. Confidence-grounded extraction is the whole point.
What about scans that are blurry, tilted or photographed on a phone?
+
This is exactly what our extractor pipeline was built for - drivers' mobile photos, scanned-and-resigned invoices, multi-page packing lists. Standard AI extractors fail on these; ours normalise orientation, locate fields and emit cited values.
Where is the data hosted?
+
EU regions by default, with optional UK hosting. We sign DPAs and configure model providers for zero retention. Self-hosted Paperless is available for clients with strict data residency requirements.
How long does a typical build take, and what does it cost?
+
A focused customs workflow MVP (ingestion + extraction + Paperless generation for one trade lane) typically ships in 4-6 weeks. Full builds with 3-way validation, CargoWise integration and the audit dashboard usually run 8-10 weeks. Build cost is typically £15-25k with an optional ongoing operate retainer.
Related automation services
Compliance document automation
Automate compliance document generation: audit packs, sub-contractor compliance, policy briefs, training records. Audit-ready in minutes, multilingual, fully owned.
Learn moreDocument assistance agents
AI agents that read your contracts, leases, customs scans and policy libraries - returning structured data and cited answers. The extraction and Q&A stage of an end-to-end document workflow.
Learn moreDocument automation
Automate generation of contracts, certificates, reports, proposals and statements from your data - at scale. REST API, CSV batches and a no-code template editor.
Learn more