AI Invoice Reconciliation & 3-Way Matching Software for ERP
Data Alchemy automates invoice reconciliation by matching every supplier invoice against the corresponding purchase order and delivery note (DDT). The platform extracts line-level data, performs three-way matching, validates the values against your ERP master data and posts the result into SAP, Zucchetti or TeamSystem — 99.8% accuracy with Claude, ~12 s for a one-page document (Claude Sonnet), no templates.
From raw invoices to reconciled ERP entries
Invoice reconciliation is the most time-consuming control in accounts payable: matching every supplier invoice with the corresponding purchase order and delivery note (DDT). Data Alchemy assigns a single AI engine to each document model — Claude by Anthropic (the engine we recommend by default), GPT by OpenAI, Data Alchemy AI (Data Alchemy's proprietary model, running on its own servers and GPUs in Italy) or a compatible LLM through the European inference providers Scaleway and Seeweb Regolo.ai, on the customer's own subscription — to read each document, extract line items and align them by SKU, quantity and price. The reconciled entry is validated against ERP master data and posted automatically, so your finance team stops re-keying invoices and reviews only genuine exceptions. Automating accounts payable means extending this control across the whole flow, from purchase order to payment.
Everything you need to reconcile invoices automatically
Line-level AI extraction
A dedicated LLM reads every invoice, purchase order and delivery note and extracts line items (item, quantity, unit of measure, price) with no templates, at 99.8% accuracy (measured with Claude).
Automatic three-way matching
Each invoice line is aligned with the matching purchase-order line and goods-received line by SKU, quantity, unit of measure and price. Discrepancies surface before posting.
ERP master-data validation
Suppliers, items, price lists and payment terms are checked in real time against SAP, Zucchetti or TeamSystem. Clean matches flow straight through.
Duplicate and anomaly detection
The AI compares each incoming invoice with the history of posted documents, blocking duplicates, double submissions and tampered totals before payment is approved.
Email-first acquisition
The Mail service, installed at your premises, manages multiple business mailboxes and sends the platform the invoices, POs and DDTs matching a model, queuing only relevant documents. Emails and credentials stay within your perimeter.
ERP checks via API
With external fields and remote validation (Business and Enterprise licences) the lines read by the AI are verified directly against your ERP: purchase order, item codes and customer code retrieved via API.
This is PO matching performed at line level, not a check on document totals: discrepancies in unit price, delivered quantity — as recorded on the delivery note or goods received note (GRN) — or item code surface before the entry reaches the ERP, so the invoice approval workflow only ever sees the lines that genuinely need a human decision.
Four steps from inbox to reconciled ERP entry
1. Capture
Invoices, purchase orders and delivery notes are captured from email, scanners or upload — PDF, XML, FatturaPA and international formats.
2. Extract with the LLM
Data Alchemy assigns a single AI engine per document model — Claude by Anthropic, GPT by OpenAI, Data Alchemy AI (Data Alchemy's proprietary model) or a compatible model through the European providers Scaleway and Regolo.ai — and extracts structured line-level data; with Claude Sonnet or higher, reading takes on average around 12 seconds for a one-page document and around 18 seconds for two pages.
3. Three-way match
Each invoice line is matched with the corresponding PO and DDT line by SKU, quantity, unit of measure and price.
4. Validate and post
Reconciled data is validated against SAP, Zucchetti or TeamSystem master data and posted into the ERP through the AI agent or the REST API.
Two-way, three-way and four-way matching: which control you need
Not every invoice needs the same level of verification. Data Alchemy automatically applies the right control based on the documents available in your management system.
Two-way matching
Compares the supplier invoice against the purchase order alone, checking SKU, ordered quantity and agreed price. It is the typical control for services and supplies with no physical delivery of goods, where no delivery note exists.
Three-way matching
Adds the delivery note: the invoice is compared with the purchase order and with the goods actually received on the DDT. It is the accounts-payable standard for goods and materials, because it is the only control that stops you paying for quantities never delivered.
Four-way matching
Also brings in the inspection report or quality acceptance record: you pay only for what was ordered, delivered and formally accepted. It is the control required in manufacturing and in purchases of plant and critical components.
Data Alchemy handles all three scenarios on the same platform: the LLM classifies incoming documents, recognises which ones are available for each order line and triggers the applicable matching level, with no separate configuration per purchase type.
Once the match clears, a second layer of verification remains on the payable cycle — duplicates, altered IBANs, amounts outside tolerance, suppliers not on file: see document anomaly control.
What happens when the three documents do not match
A real accounts-payable cycle never matches at 100%. The validation rules, configured with you during onboarding, decide line by line what passes automatically and what stops for human review.
Price tolerance
Accept a deviation from the order price within a threshold you define, as a percentage or an absolute value. Anything outside the range is flagged before posting.
Quantity tolerance
Handle over- and under-deliveries by absolute quantity or percentage, and let the platform decide whether to accept, partially match or hold the line.
Partial and split deliveries
One purchase order can be closed by several delivery notes, and one invoice can cover several DDTs. Data Alchemy reconstructs the many-to-many link between order lines, delivered quantities and invoiced quantities instead of forcing a one-to-one document match.
Ancillary charges
Transport, packaging, surcharges and rounding present on the invoice but not on the order are recognised and routed to the correct accounting rule rather than treated as a mismatch.
Exception routing
Every line that fails the check is queued to the responsible reviewer — buyer, warehouse or accounting — with the three source documents side by side, so the exception is resolved by reading rather than rebuilt from scratch.
The discrepancies three-way matching catches
Three-way matching compares three documents — supplier invoice, purchase order and delivery note (DDT) — before approving payment. Data Alchemy automatically surfaces the deviations that would otherwise require manual checks.
Price mismatch
The invoice price does not match the agreed order price or price list: the exception is flagged before posting.
Under-delivered quantity
The invoiced quantity exceeds what was actually received on the DDT: the line stays on hold until verified.
Duplicate invoice
Same supplier, same number or amount already posted: the system blocks the double payment.
Off-catalogue item
An item code missing from the ERP master data is highlighted for coding or approval.
The delivery note checked line by line against the order in your ERP
Three-way matching works best when the purchase order does not have to be rebuilt from a PDF but is read straight from the management system. The Business and Enterprise licences include two capabilities that connect extraction to your ERP.
Verifying AI-read data against external services
When reading a delivery note that references a purchase order, Data Alchemy queries the customer's ERP via API, retrieves the order contents and checks item code, quantity and the other fields line by line: the data read by the AI is consistent with the data in the management system.
External fields →Validation against a third-party source such as your ERP
Remote validation connects a third-party source such as your ERP to check item codes and retrieve the customer code from the company name or VAT number: the line that reaches matching already uses your ERP's codes.
Remote validation →External fields and remote validation are included in the Business and Enterprise licences. Compare the licences on the pricing and licences page.
Built for finance, logistics and procurement teams
Accounts payable automation
The end-to-end order-DDT-invoice process: from reconciliation to posting, with the ROI for the finance team.
Learn more →Finance & APAP automation by department
Reconcile supplier invoices with POs and DDTs, validate against the ERP and post entries with no manual re-keying.
Learn more →GPT reconciliationGPT invoice reconciliation
See the LLM-powered angle: how GPT, chosen as the engine for your invoice model, drives zero-template 3-way matching.
Learn more →IDP PlatformFull IDP solution
Beyond invoices: orders, contracts, price lists and bank statements processed by the same AI platform.
Learn more →Logistics industryAutomatic coding in logistics
Delivery note, CMR and carrier invoice lines coded against ERP master data, with only exceptions routed to a reviewer.
Learn more →Frequently asked questions about invoice reconciliation
What is automated invoice reconciliation?
Automated invoice reconciliation uses artificial intelligence to read supplier invoices, match each line against the corresponding purchase order and delivery note (DDT), and post the result into the ERP. Data Alchemy assigns a dedicated LLM to each document model — Claude by Anthropic, GPT by OpenAI, Data Alchemy AI (Data Alchemy's proprietary model) or a compatible model through the European providers Scaleway and Regolo.ai — and, with Claude, reaches 99.8% extraction accuracy (no accuracy figure is published for the other engines) and 3-second processing per document, without templates.
What is three-way matching (invoice-order-DDT)?
Three-way matching is the standard accounts-payable control that compares three documents — supplier invoice, purchase order and delivery note (DDT) — before approving payment. Data Alchemy automates this control: the LLM extracts line items from each document, the platform aligns them by SKU, quantity and price, and the system flags discrepancies (price mismatch, missing goods, duplicate invoice) before the entry hits the ERP.
How is it different from traditional OCR reconciliation?
Traditional OCR converts pixels into text but cannot decide whether an invoice line belongs to a specific purchase order or whether quantities match the delivery note. An LLM-powered reconciliation engine understands document semantics: it identifies suppliers, items, units of measure and prices, then performs three-way matching against the ERP master data. The result is structured, validated and ready to post — no human re-keying required.
Does it integrate with SAP, Zucchetti and TeamSystem?
Yes. Reconciled invoices flow directly into SAP, Zucchetti, TeamSystem and other ERPs through the AI agent installed at your premises (a database, CSV/JSON/TXT files or APIs) or with an integration built on Data Alchemy's REST API and webhooks. Master data (suppliers, items, price lists) is read from the ERP in real time so the AI can validate against the system of record on every document.
Which invoice formats does the platform handle?
The platform processes PDF, XML, FatturaPA and international electronic invoice formats, as well as scanned paper invoices. It also ingests purchase orders and delivery notes from email (business mailboxes, via the on-premise Mail service), scanners or upload, classifying each document automatically before reconciliation.
How much does AI invoice reconciliation cost?
Data Alchemy works on a licence model — Basic, Business and Enterprise — plus pay-as-you-go credits, with a quote on request; to try the platform you start from the Demo (a trial Basic licence). External fields and remote validation, useful for checking delivery notes and invoices against the order in your ERP, are included in the Business and Enterprise licences. The licence comparison is on the pricing page.
Can Data Alchemy check a delivery note against the purchase order in my ERP?
Yes, with external fields, included in the Business and Enterprise licences. When reading a delivery note that references a purchase order, Data Alchemy queries the customer's ERP via API, retrieves the order contents and checks item code, quantity and the other fields line by line. Remote validation completes the check by verifying item codes and retrieving the customer code from the company name or VAT number.
See invoice reconciliation in action
Activate the free Demo with 10 documents and check the AI reading on your real invoices, with 1 model chosen from the predefined ones in a dedicated environment. To see the three-way match against your POs and DDTs and calculate the ROI for your case, contact us to plan a call with a demo on your own documents.