Skip to content
SLM-Works

Available

SLM-Works Invoice Generator

Structured invoice creation from unstructured input

Turns messy source content into invoice-ready structured payloads. It focuses on extraction consistency, validation hints, and mapping fields to downstream finance workflows.

3B~92% fewer tokens per invoice vs GPT-4

How it works

  1. Step 1

    Parse source documents (email body, attachment text, OCR output).

  2. Step 2

    Extract supplier identity, line items, terms, VAT/tax, and totals.

  3. Step 3

    Score confidence per field and flag missing mandatory invoice metadata.

  4. Step 4

    Return normalized JSON that your ERP or AP workflow can ingest directly.

Example

Example input

Email text + attached PDF with free-form items, discounts, and mixed VAT percentages.

Example output

{ vendor: 'Acme BV', invoice_number: 'INV-2026-044', currency: 'EUR', line_items: [...], tax_breakdown: [...], payment_terms: '30 days' }

Key features

  • Line-item extraction with quantity/price consistency checks
  • Tax-rule aware output for multi-rate invoices
  • JSON schema validation for finance integrations
  • Confidence scores to support human-in-the-loop approvals

Rollout guidance

  • Best rollout path: start with one vendor cluster, then expand templates.
  • Pair with approval thresholds to auto-post only high-confidence invoices.

Ideal for

Finance teamsAccounting firmsERP integratorsProcurement departments

FAQ

Can this replace manual AP checks immediately?

Usually no. Teams start with assisted review, then automate straight-through processing above agreed confidence thresholds.

Want this model in your stack?

We can scope a deployment blueprint, evaluation set, and integration plan for your data and infrastructure constraints.