Ashford Port Health Authority
Using AI to unlock data trapped in import documentation, cutting compliance processing time in half and doubling team productivity in week one.
Organisation overview
Ashford Port Health Authority is a UK port health authority responsible for the inspection and control of goods of animal origin and regulated plant products entering Great Britain through the Port of Dover and Eurotunnel. The authority operates under the Border Target Operating Model, performing documentary, identity and physical checks on consignments covered by IPAFFS, the Import of Products, Animals, Food and Feed System.
The challenge
Border health documentation is inherently complex. Each consignment arriving at a border control point must be accompanied by specific certificates: Common Health Entry Documents (CHEDs), phytosanitary certificates, export health certificates and supporting declarations. Each in a different format, from a different issuing authority, containing critical fields that must be validated before a consignment is cleared or held.
At Ashford, inspectors were processing this documentation manually. Documents arrived as PDFs or scanned files, with no standard structure and no automated way to extract the data fields that determine whether a consignment passes or requires further examination. Each inspection required an inspector to open multiple documents, locate the relevant fields and manually cross-reference them against the consignment record. The process was accurate but slow, creating a throughput constraint at one of the UK's busiest border control points.
The challenge was not a shortage of data. The data existed in the documents. The challenge was unlocking it: automatically, reliably and at the volume required to keep goods moving.
Our approach
Cospire designed and delivered a solution using AI to extract structured data from unstructured border documentation. Rather than treating document processing as a peripheral workflow, we positioned it as the primary integration challenge: the moment at which paper or digital documents become structured, actionable data that Salesforce can act on.
Our Sensor accelerator handles the intake layer, converting complex, multi-format documents to AI friendly formats while preserving the structural layout, tables, headers and hierarchical content that border documentation relies on. Once in a consistent structured format, AI models extract the critical data fields, classify the document type and validate the extracted values against expected ranges and regulatory requirements.
Exceptions, cases where confidence is below threshold, or where extracted values trigger a regulatory flag, are routed to a human inspector for review. This human-in-the-loop design was deliberate: the system amplifies inspector capacity without removing inspector judgement from the compliance decision.
MuleSoft provides the integration layer, connecting the AI extraction pipeline to the Salesforce case management environment and the IPAFFS border management system. The result is a consignment record in Salesforce that is populated with verified, AI-extracted document data, ready for the inspector to review and act on.
Phase 1 was delivered in 4 months from project initiation to go-live.
The outcome
The solution transformed a manually intensive compliance process into an AI-assisted workflow. Inspectors who previously spent significant time extracting and cross-referencing data from documents now begin their review from a pre-populated, validated case record. The time saved per consignment is compounded across the volume of consignments processed each day, producing a step-change in throughput without any increase in headcount.
Technology used
- Sensor accelerator - Converts complex documentation into AI friendly format
- MuleSoft Anypoint - API-led integration connecting document pipeline to Salesforce and TRACES
- MuleSoft Intelligent Document Processing - AI powered text extraction, converting unstructured data into structured records
- Salesforce Platform - Case management, inspection record and compliance audit trail
- Human-in-the-loop review workflow - Exception routing for low-confidence extractions