Document AI & Back-Office Automation

Six EU regulations. Five new document formats. One extraction engine.

XRechnung, Factur-X, KSeF XML, CBAM emissions reports, ESG questionnaires, REACH declarations — extracted, validated, and posted to your ERP. Works with SAP, Odoo, NetSuite, Sage. Live in 14 days.

We'll show you what we'd extract from your actual document — confidence scores, ERP-ready JSON, pixel-linked audit trail. One email, one round-trip.

The pain

Three regulations that just landed on your procurement desk.

None of them existed in their current form 24 months ago. All of them turn PDFs into non-negotiable obligations — enforced on a quarterly clock, with fines as the fallback.

01 · E-invoicing

Five mandates. Five formats. One inbox.

Germany mandatory receipt since Jan 2025. Belgium PEPPOL since Jan 2026. Poland KSeF rolling through April 2026. France all-receipt September 2026. Italy SdI since 2019. Buy cross-border and you're now ingesting XRechnung, Factur-X, KSeF XML, UBL 2.1 and PEPPOL BIS in parallel — each with its own quirks.

Sources · European Commission · EY · Avalara

02 · CBAM

Every import is now a supplier emissions interrogation.

The CBAM definitive period started 1 Jan 2026. Iron/steel, aluminium, cement, fertilizer, hydrogen importers must chase verified installation-level emissions from non-EU suppliers — or pay punitive default values. First verified annual report due 30 Sep 2027, and half your suppliers send PDFs in Mandarin or Turkish.

Sources · EC Taxation & Customs · Coolset

03 · ESG questionnaires

Your big customers send 80-question surveys. Every quarter.

CSRD reporting cascades down through every value chain. Even with the VSME cap (EC recommendation July 2025), most large buyers still send bespoke non-VSME questionnaires. Your ops person reopens a spreadsheet she filled in three months ago and can't remember which answers she already gave.

Sources · EFRAG · European Commission

04 · REACH / SVHC

SVHC candidate list updates 2×/year. Every update re-opens every supplier file.

05 · ViDA

VAT in the Digital Age compresses AP cycles from 3 weeks to days. Your manual process dies first.

The anchor build

E-Invoice & AP Automation

Every inbound invoice — structured XML or PDF scan — parsed, validated, matched against your POs, and posted to your ERP. Before your AP clerk opens Outlook.

What it does
  • Ingests PDF, XRechnung, ZUGFeRD, Factur-X, KSeF XML, PEPPOL BIS — same pipeline
  • 3-way match: PO · invoice · goods receipt — flags mismatches for review
  • Validates VAT rates per EU country (DE 19%, FR 20%, IT 22%, etc.)
  • Pushes to SAP, Odoo, NetSuite, Sage via API — sandbox test first, never your prod
  • Confidence score on every field, pixel-link back to source PDF — audit-ready
  • Low-confidence fields route to a human reviewer; clean ones go straight to the ERP
Stack
ClaudePEPPOLSAP / Odoo / NetSuite / SagePostgresn8n
Typical outcome

85–95% of invoices posted with zero human touch.

The rest route to a review queue with the exact field flagged. 14-day pilot on your inbox — cancel per SOW terms.

Live demo

Watch an XRechnung get parsed and posted — 3.4 seconds.

Press play. Fields stream in with confidence scores exactly the way the production pipeline writes them. For an extraction on your own document, send us a PDF.

Live demo · sample extraction
E-Invoice → SAP in 3 seconds
Idle
Press Start extraction to watch a real XRechnung invoice get parsed, validated, and posted to an ERP — field by field, with confidence.
Send your own PDF →
What just happened

Parsed, validated, matched, posted — in 3.4 seconds.

  • Detected XRechnung XML · normalised to internal schema.
  • VAT rate validated against DE standard (19%) — pass.
  • 3-way match: PO #4501 · GR-2026-0412 · invoice total — pass.
  • Low-confidence fields: none. Would route to human if <95%.
  • Posted to SAP S/4HANA. Source PDF pixel-linked in audit trail.
This is a scripted demo on a real XRechnung sample. Your version runs on your own documents, posts to your ERP, and is live in 14 days. Send a sample PDF to see it extract from a real one of yours.

Proof

We've built this engine for food co-packers.

The extraction stack — detect format, extract fields, validate, confidence-score, human-in-the-loop on low confidence, push to ERP — ports cleanly from brand RFPs to CBAM emissions reports to ESG questionnaires to REACH declarations. Same pipeline, different schema.

Also available in this niche

Document AI isn't just invoices.

01

CBAM Supplier Emissions Extractor

Parses multilingual supplier emissions PDFs into the CBAM reporting template — CN codes, installation IDs, verifier accreditations, all validated against the quarterly schema.

02

ESG Questionnaire Auto-Fill

Builds an answer-memory from every CSRD cascade and Ecovadis survey your big customers send. Auto-fills the ~70% overlap and flags out-of-scope items against the VSME cap.

03

REACH / SVHC Compliance Monitor

Ingests supplier REACH Art. 33 declarations, diffs every SVHC candidate-list update (2×/year), and re-verifies affected SKUs automatically — no more day-per-supplier onboarding.

Objections we hear

The seven questions every procurement director asks.

Will it break our SAP / Odoo / NetSuite / Sage?+

No. We write into the ERP through the same API your team already uses, or CSV import if you prefer. Never modify the schema. First integration test runs on a sandbox copy before anything touches production.

Where does the document go? Who processes it? GDPR?+

EU-region infrastructure only (Frankfurt or Paris). Documents are never used for model training. We sign a GDPR Art. 28 processor agreement with a named sub-processor list before the first upload.

What happens when the AI extracts a wrong field? We get audited.+

Every extraction has a confidence score and a pixel-link back to the exact region in the source PDF. Below-threshold fields route to a human queue. Every change is logged with user, timestamp, and before/after — that's your audit trail.

Our AP clerks ignored the last RPA rollout. Why would this be different?+

We don't ship a new app to log into. Output lands in the tool they already use — Outlook approvals, the ERP inbox, or Teams. We measure adoption in the first 30 days and iterate on the workflow, not the tool.

If we want to leave in 18 months, are we hostage?+

Your field schemas, prompt configs, and processed-document history export to JSON/CSV on request. No proprietary model weights you can't take. The extraction prompts are yours in plain text.

We already have ABBYY / Rossum / Kofax / a Python OCR script. Why you?+

Keep it. Our wedge is the messy, low-volume, high-variance docs where template-based OCR fails today — CBAM emissions PDFs, REACH attestations, ESG questionnaires. We plug in alongside your existing stack. We don't replace it.

Enterprise Document AI starts at €30K/year. How can you do this at €500/month?+

Productized flat fee for a defined document scope, plus the Founding 5 rate locking the first 5 Matchquote clients across all services at €0 setup + €500/month. You're not paying for a sales team, a gold-tier SLA, or 200 features you'll never use. Same engine, smaller blast radius.

Price

€0 setup + €500/month.

Founding 5 rate — first 5 Matchquote clients across all services. Locked in forever.

14-day pilot on your real inbox. Cancel per SOW terms. Client #6+ pays €1,500 setup + €1,500/month.

  • 14-day pilot
  • EU-region
  • Art. 28 DPA
  • Cancel per SOW terms