IDP Technology: the AI That Reads, Validates and Posts Your Documents
Every extracted data point is verified in real time against your management system's master data. The system learns from your confirmations and continuously improves.
This extraction and validation technology is the engine behind the Data Alchemy Intelligent Document Processing (IDP) platform.
What is Intelligent Document Processing (IDP) and why choose Italian IDP solutions
Data Alchemy's Intelligent Document Processing (IDP) is document data extraction software that automates the reading, validation and ERP posting of business documents: supplier invoices, FatturaPA XML electronic invoices, delivery notes (DDT), customer orders, price lists and contracts. Unlike traditional OCR software, which only converts an image into text, Data Alchemy assigns a dedicated LLM to each document model: the AI understands what it is reading, cross-checks every field against your ERP master data in real time and posts the finished record with 99.8% end-to-end accuracy (measured with Claude) and no templates to build or maintain; with Claude Sonnet or higher, reading takes on average around 12 seconds for a one-page document and around 18 seconds for two pages.
Why choose IDP software over traditional OCR?
- ✓Contextual understanding: AI interprets the meaning of data, not just reads it
- ✓Automatic validation: every field is verified against ERP master data in real time
- ✓Continuous learning: the system improves with every processed document
- ✓Zero configuration: no templates to create, works immediately on any format
The four checkmarks above summarise the difference, but the comparison is clearest side by side.
IDP vs OCR vs template-based capture: the real differences
| Criterion | Traditional OCR | Template-based capture | Data Alchemy IDP |
|---|---|---|---|
| What it does | Converts the image into raw text | Reads fixed coordinates on a known layout | Understands the meaning of the document |
| New supplier layout | Text extracted, fields not recognised | Requires a new template (days of work) | Works immediately, zero configuration |
| Handwriting, stamps, crooked scans | High error rate | Template misaligned, extraction fails | Interpreted in context by the LLM |
| Data validation | None | None or basic format checks | Every field cross-checked against live ERP master data |
| Maintenance | None, but constant manual correction | One template per supplier, forever | None: the model learns from your confirmations |
| Typical accuracy | 60–80% on real fields | 90–95% only on mapped layouts | 99.8% end-to-end with Claude |
| Output | Unstructured text file | Structured record, no validation | Validated record posted straight into SAP, Zucchetti or TeamSystem |
In practice: with OCR someone still has to read the text and type it in. With templates, every new supplier means a new configuration project. With Data Alchemy's IDP software, a document from a supplier you have never seen before is processed correctly the first time.
IDP Solutions for Italian companies
Data Alchemy's IDP (Intelligent Document Processing) solutions address the specific needs of Italian SMEs: handling of XML FatturaPA electronic invoices, paper and digital DDTs, customer orders via email and commercial contracts. Unlike generic OCR software, our platform understands the Italian document context, automatically recognises VAT numbers, fiscal codes, SDI recipient codes and VAT rates, and integrates natively with the major ERP systems used in the Italian market such as SAP Business One, Zucchetti Ad Hoc, TeamSystem Enterprise and Mago.net.
Real-time verification
Every field is cross-checked against ERP/CRM master data: suppliers, items, prices, conditions. Anomalies are flagged before posting.
Automatic enrichment
Missing data in the document is automatically completed by pulling from master data: item codes, ledger accounts, cost centers.
Learning after 3 matches
After 3 matches confirmed by the same operator, the system automatically suggests future correspondences. Less human intervention, more speed.
Exception handling
Documents with uncertain data are routed to the appropriate operator with all data pre-filled. One click to confirm or correct.
How IDP software reduces manual data entry
Every supplier invoice, delivery note or customer order keyed in by hand costs the finance team time and introduces errors that only surface at reconciliation. Data Alchemy's IDP software removes the typing: the LLM assigned to that document model reads the document, extracts the fields (supplier, VAT number, document number and date, line items, taxable amount, VAT rates, total) and checks them against the ERP master data before posting.
What actually changes in day-to-day work:
- ✓No line-by-line typing: data arrives already structured and ready to be posted into the ERP.
- ✓Anomalies caught before posting: prices outside agreed terms, unknown suppliers, duplicate documents.
- ✓Operators move from data entry to exception handling: they only touch the documents the AI flags as uncertain.
- ✓No templates to build or maintain when a supplier changes the layout of an invoice or delivery note.
The result is an accounts payable cycle that scales with volume without new hires: document automation absorbs the month-end peaks while the team stays focused on reconciliation, due dates and supplier relationships.
From here there are two concrete next steps: the guide on how to automate accounts payable walks through the full order-to-payment flow, and the free Demo on your own documents shows what the AI actually extracts from them.
Data Alchemy's AI engines: Claude, GPT, Data Alchemy AI and European providers
For each document model you choose one AI engine: Claude (Anthropic), GPT (OpenAI), Data Alchemy AI, Data Alchemy's proprietary model, or a compatible LLM through the European inference providers Scaleway and Seeweb Regolo.ai. Claude is the engine we recommend by default.
For each document model you choose one AI engine: Claude (Anthropic), the engine we recommend by default; GPT (OpenAI); Data Alchemy AI, Data Alchemy's proprietary model, running on Data Alchemy's own servers and GPUs in Italy; or a compatible LLM through the European inference providers Scaleway (Generative APIs) and Seeweb Regolo.ai, on a subscription taken out directly by the customer. Each document is processed by a single engine, the one paired with its model: for example the invoice model on Claude, the delivery note (DDT) model on GPT and the order model on Data Alchemy AI, or several models on the same engine.
Data Alchemy's AI engine: one LLM per document model
At the heart of Data Alchemy's IDP software is a simple principle: each document model — invoices, delivery notes, orders, contracts — is paired with a dedicated LLM that processes all documents of that model. No templates, no manual rules: the model understands the content and extracts the data with 99.8% end-to-end accuracy on real-world documents, measured with Claude.
One AI engine for each document model
For each document type you configure in Data Alchemy — invoices, delivery notes, orders, contracts — you choose a single AI engine that handles all documents of that model: for example the invoice model on Claude, the delivery note (DDT) model on GPT and the order model on Data Alchemy AI, or several models on the same engine. Extraction is therefore consistent and specialised for that type, with no templates or manual rules.
Claude AI (Anthropic)
Anthropic's Claude AI is the engine we recommend by default: we recommend it for its accuracy, instruction following and structured-output reliability on real business documents. It excels at contextual understanding of long, nuanced text — contracts, payment terms, non-standard clauses — and at reading scans, photos and complex layouts, with a low tendency to hallucinate data.
GPT (OpenAI): available
OpenAI's GPT is available as an AI engine for your document models: you can choose it for one or more models, as an alternative to Claude, Data Alchemy AI and the European inference providers. With GPT too, each document is processed by a single engine and the extracted data is validated against ERP master data.
Data Alchemy AI, our proprietary model: available now
Data Alchemy AI, Data Alchemy's proprietary model, is available as an alternative to third-party engines and runs on Data Alchemy's own servers and GPUs in Italy. You can choose it for one or more document models, just like the other engines.
European inference providers: Scaleway and Seeweb Regolo.ai
If you prefer inference on third-party European infrastructure, you can take out your own subscription with Scaleway (Generative APIs, data centres in France) or Seeweb Regolo.ai (data centres in Italy), choose a compatible LLM and use it as the AI engine inside Data Alchemy for one or more document models. Here too each document is processed by a single engine and the extracted data is validated against ERP master data.
AI agent, REST API and webhooks: integration with SAP, Zucchetti and TeamSystem
Data Alchemy's IDP platform integrates with any ERP or management system through the AI agent installed at your premises (a database, CSV, JSON or TXT files, or APIs) or Data Alchemy's REST API and webhooks, with full support for the Italian electronic invoicing standard (FatturaPA). No vendor lock-in, maximum flexibility, ready in days, not months.
See how native ERP integration works and what it means for your business system.
Integration Methods
Four ways to connect: REST API and webhooks for developers, the AI agent installed at your premises for databases, files or APIs, the native Salesforce connector and the Mail service.
Output Formats
Extracted data is returned in the format best suited to your workflow.
Compatible ERPs
Native integration with the most widely used ERPs in businesses.
CRM
Native Salesforce integration, included in the Enterprise licence: data extracted from orders and documents reaches your CRM with no manual data entry. Other CRMs connect through the AI agent or the REST API.
Email Integration
The Mail service, installed at your premises, manages multiple business mailboxes from a graphical interface and sends Data Alchemy only the emails matching the assigned model. Emails and credentials stay within your perimeter. Available from the Business licence.
Security and Access
SSO with Microsoft Entra ID and Google Workspace via LoginMaster, role-based permissions and compliance with European regulations.
Full ERP/CRM integration
Automate document acquisition from your mailboxes with the Mail service
The Mail service is installed at the customer's premises (on-premise): emails, which are sensitive data, and mailbox credentials stay within the company perimeter and not on Data Alchemy's cloud servers. It manages multiple business mailboxes, is configured from a graphical interface and sends Data Alchemy only the emails matching a document model. It is available in the Business and Enterprise licences.
How it works
- 01Install the Mail service: it runs in your infrastructure, so mailboxes and credentials stay within your perimeter
- 02Configure mailboxes and models from the graphical interface: pair a dedicated address (e.g. logistics@) with a specific document model such as delivery notes, invoices or orders
- 03Routing by model: only emails matching a model are sent to Data Alchemy; spam, newsletters and unrelated mail stay in the mailbox
- 04Automatic processing: relevant documents (PDF attachments, images, structured emails) are extracted and processed just like any manually uploaded document
Practical example
From the Mail service interface you pair logistics@ with the delivery note model. A supplier sends a delivery note PDF by email: the Mail service sends it to Data Alchemy, which processes it and enters the data into your ERP. If a newsletter or marketing email lands in the same mailbox, it matches no model and stays in the mailbox.
On-premise
The Mail service is installed at your premises: emails and mailbox credentials stay within your perimeter, not on Data Alchemy's cloud servers.
Graphical interface
Mailboxes and models are configured from a graphical interface, with no scripts or configuration files. Available in the Business and Enterprise licences.
Routing by model
Only emails matching the configured model are sent to Data Alchemy: spam and unrelated mail stay in the mailbox.
Multi-mailbox
A single Mail service manages multiple mailboxes paired with different models: one for delivery notes, one for invoices, one for orders.
Enterprise authentication and a dedicated authentication tenant for every client
Data Alchemy supports login with Microsoft Entra ID and Google Workspace through LoginMaster, with a dedicated authentication tenant for each client. Enterprise users from Entra and Google Workspace are paired with native Data Alchemy users: Single Sign-On (SSO) access and, on the platform, logical separation of client data, in line with data sovereignty in Europe. See our security & compliance page for details on GDPR, EU hosting, AI processing and encryption in transit.
SSO with Microsoft Entra ID and Google Workspace
Enterprise Single Sign-On with Microsoft Entra or Google Workspace credentials (via LoginMaster), with no separate passwords to manage.
Entra + Google Workspace + native
Microsoft Entra and Google Workspace users are paired with native Data Alchemy users.
Authentication tenant per client
Each client gets its own authentication tenant on LoginMaster; on the platform, client data is logically separated.