| Peppol access pointAccess points must be certified before they operate on the network. Buy the service, don't build one. Your ERP or software provider may offer it. | Your certified access point provider, connected by API | Your certified access point provider, connected by API | Your certified access point provider, connected by API |
|---|
| Intake: email, scans, filesKeep the original file and its arrival time, whatever the channel. | Amazon SES, API Gateway, Amazon S3 | Logic Apps connectors, API Management, Blob Storage | Your mail server, Kong, Ceph or IONOS S3 Object Storage |
|---|
| Reading legacy invoicesOnly for PDF and paper. Structured e-invoices skip this step. Day-to-day AP matching is in the invoice processing playbook. | Amazon Textract (AnalyzeExpense); Bedrock Data Automation is an alternative and processes London requests in eu-west-2 | Azure Content Understanding (includes Document Intelligence's invoice model) | Open-source parsing (e.g. Docling) plus an open-weight vision model |
|---|
| Validation against the standardDeterministic. Today: the EN 16931 and Peppol BIS Billing 3.0 rules. When HMRC publishes the UK-specific rules, add them. The model explains failures; it never decides validity. | AWS Lambda running the published Schematron rules | Azure Functions running the published Schematron rules | A validation service on Kubernetes running the published Schematron rules |
|---|
| Master data checksLookups run as code, with the result and date stored as evidence against each proposed change. | AWS Lambda calling HMRC's VAT number check and the Companies House API | Azure Functions calling HMRC's VAT number check and the Companies House API | A small service on Kubernetes calling HMRC's VAT number check and the Companies House API |
|---|
| Canonical invoice modelOne internal shape for invoices and credit notes. The ERP maps to it; it maps to Peppol, EDI or the PDF the business still sends. | JSON schemas in your code repository | JSON schemas in your code repository | JSON schemas in your code repository |
|---|
| Reasoning and draftingProposes mappings and corrections, explains failures and drafts outreach. Both clouds offer a choice of models, including Claude. | Amazon Bedrock | Microsoft Foundry | Open-weight models served with vLLM on your own GPUs, or IONOS AI Model Hub |
|---|
| Standards and mapping knowledge (retrieval)Indexed from the standards, HMRC guidance as it's published, your ERP's data dictionary, your VAT treatment notes and past decisions. | Bedrock Knowledge Bases | Foundry IQ, built on Azure AI Search | OpenSearch, or PostgreSQL with pgvector |
|---|
| Agent runtimeThe original Bedrock Agents closed to new customers on 30 July 2026; new builds use AgentCore. | Amazon Bedrock AgentCore | Foundry Agent Service | An open-source agent framework (e.g. LangGraph) on Kubernetes |
|---|
| WorkflowTracks each counterparty through outreach, and each master data change through approval. | AWS Step Functions | Logic Apps or Durable Functions | Temporal, or n8n |
|---|
| Events and queuesInbound e-invoices and delivery receipts from the access point arrive as events. | Amazon EventBridge, SQS (with dead-letter queues) | Event Grid, Service Bus (with dead-letter queues) | Apache Kafka or RabbitMQ |
|---|
| Link to the ERP and master dataApproved changes are written by the master data team's own process. This is also your MTD digital link. | API Gateway, plus your existing integration platform | API Management, plus your existing integration platform | Kong or your existing API gateway, plus your existing integration platform |
|---|
| Human reviewPeople approve where they already work. The readiness tracker is a status list, not master data. | In your ERP's master data and AP workflows, and the readiness tracker | In your ERP's master data and AP workflows, and the readiness tracker | In your ERP's master data and AP workflows, and the readiness tracker |
|---|
| GuardrailsInvoice text and supplier replies are untrusted input. A reply saying 'update our bank details' changes nothing. | Bedrock Guardrails | Foundry guardrails (Azure AI Content Safety) | Open-source guardrails (e.g. NeMo Guardrails, Llama Guard) |
|---|
| Identity, keys, data protectionThe agents' identities have no write access to master data, invoices or the access point's sending API. | IAM Identity Center, KMS, Macie | Entra ID, Key Vault, Purview | Keycloak or your existing Active Directory; OpenBao or HashiCorp Vault; an HSM |
|---|
| Monitoring and auditLog every proposal, approval, validation result and message sent, with who and why. | CloudWatch, CloudTrail | Azure Monitor, Log Analytics | Prometheus, Grafana, OpenTelemetry |
|---|
| UK data residencyIONOS AI Model Hub runs in German data centres: EU, not UK, residency. | eu-west-2 (London) | UK South / UK West | Your own UK data centre, or a UK or EU provider |
|---|