The IDP platform: from documents to your ERP, no data entry
The Intelligent Document Processing platform by Data Alchemy covers the document's entire journey: it captures invoices, delivery notes, orders, contracts and price lists, extracts the data with AI, validates it against your ERP master data and writes it straight into the ERP. A single end-to-end flow that eliminates manual data entry and takes the document from the inbox to a posted record in your ERP; with Claude Sonnet or higher, reading takes on average around 12 seconds for a one-page document and around 18 seconds for two pages.
accuracy with Claude
Per 1-page document (Claude Sonnet)
templates to configure or maintain
natively integrated ERPs
The end-to-end flow: capture, extraction, validation, writing
The platform is organized into four stages that run automatically one after another. Each stage is designed to work without manual intervention, with checks that guarantee only correct data reaches the ERP.
Capture
Documents enter automatically from email, monitored folders or the API.
Documents come in as PDFs, images, scans or FatturaPA XML: you can upload them via folder or send them to the API. For email, the Business and Enterprise licences include the Mail service, installed at your premises: it manages multiple business mailboxes from a graphical interface and sends Data Alchemy the emails matching a model (invoices, orders, delivery notes), while emails and mailbox credentials stay within your perimeter, not on Data Alchemy's cloud servers. No more forwarding invoices to the back office or downloading them by hand: they enter the flow on their own.
Multi-engine AI extraction
An LLM understands the document and extracts the right fields, with no templates.
For each document model you choose one AI engine: Claude (recommended), GPT, Data Alchemy AI or a compatible model through the European providers Scaleway and Regolo.ai; each document is read by the engine paired with its model. The AI understands the content and pinpoints taxable amount, VAT, due dates, supplier and line items even on layouts it has never seen, with no templates to configure. It automatically recognizes invoices, delivery notes, orders, contracts and price lists.
Validation
Data is checked against ERP master data and business rules.
Before writing, every extracted field is validated in real time against the ERP master data: suppliers, items, VAT rates, terms. The platform flags anomalies and inconsistencies — an unknown supplier, a total that doesn't add up — so questionable documents are handled before they enter the management system, not after. With external fields and remote validation (Business and Enterprise licences) the check extends to external sources such as your ERP, queried via API.
Writing into the ERP
Validated data is written into the ERP through the AI agent or the REST API and webhooks.
The final step writes the structured data into the ERP (SAP, Zucchetti, TeamSystem, Microsoft Dynamics 365, Oracle NetSuite or others) 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. The document is posted, not just digitized: the operator has nothing to re-enter.
The result is a continuous cycle: the document enters from the inbox and leaves as a posted record in the management system, with a full audit trail of every step.
The platform's technical differentiators
Beyond the flow, what sets the platform apart is how it's built: an architecture designed for accuracy, integration and scalability across the document cycle of Italian SMEs.
Multi-engine approach
Not one model for everything. 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. 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. Each document is processed by a single engine at a time: the one chosen for its model.
Template-free extraction
No per-supplier or per-layout configuration: the AI understands the document and finds the right fields on the first pass, even on formats it has never seen. No templates to build and no maintenance at every invoice change.
Validation against ERP master data
Validation isn't a simple format check: it compares the extracted data with the real master data in the management system and with business rules, blocking errors before they reach the ERP.
Native ERP integration
Connection to SAP, Zucchetti, TeamSystem, Microsoft Dynamics 365, Oracle NetSuite and AS400 (IBM i) 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. Data arrives in the management system in the format it expects.
Developer API
Beyond email capture, the platform exposes a document extraction API: you send a document and receive structured JSON, to embed extraction into existing flows and applications.
Native FatturaPA support
The platform handles FatturaPA XML documents and Italian tax formats: it's tailored to the accounts-payable and receivable cycle of Italian companies, not a generic IDP to adapt.
External fields and remote validation: AI-read data checked against your ERP
Beyond master-data validation, the Business and Enterprise licences include two capabilities that connect extraction to external sources: the data read by the AI is not just format-checked, it is verified against what your management system already knows.
Verifying AI-read data against external services
External fields enrich and verify the extracted fields by querying external services, for example your ERP via API.
Example: 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.
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.
Example: An order arrives with only the customer's company name and VAT number: remote validation queries the management system, retrieves the matching customer code and checks that the item codes on the order really exist in master data.
One platform, three licences and a free trial
The platform is the same for everyone: what changes is what each licence includes. You start with the Demo, the free trial of the Basic licence, then choose the licence that fits your needs and use pay-as-you-go credits, with a quote on request.
Basic — free trial (Demo)
The Basic licence on trial, to see the platform on your own documents.
- Available soon
- Free for 10 days, 10 documents
- 1 model chosen from the predefined ones, with the storage it needs
- Dedicated environment, data never shared
- AI agent not included
Basic
To get started with one document flow.
- 1 model
- 500 MB of storage, with a mandatory 7-day retention: documents are deleted automatically 7 days after upload
- Pay-as-you-go credits, bought online
- AI agent not available
Business
Everything in Basic, plus mail and ERP integration.
- On-premise Mail service
- External fields
- Remote validation
- Models and storage sized to your needs
- Files kept for up to 60 days from upload, configurable per model
- Credits in annual quantities, lower unit price by quantity
- AI agent available as a separately quoted add-on
Enterprise
Everything in Business, plus CRM, Order Dispatcher and the AI agent.
- Native Salesforce integration (CRM)
- Order Dispatcher: orders land as drafts in your CRM, today through the native Salesforce connector; connector architecture designed to extend to other management systems
- AI agent included
- Tailored models and storage
- Credits in annual quantities, lower unit price by quantity
Which documents the platform processes
A single platform for all your business documents: each one is recognized automatically, extracted and written into the correct ERP module.
Invoices
Supplier and customer invoices, including FatturaPA XML: taxable amount, VAT, due dates and line items ready to post.
Delivery notes
Delivery notes (DDT): packages, quantities, order references, for reconciliation with invoices and orders.
Orders
Purchase and sales orders: items, quantities, prices and terms written straight into the management system.
Contracts
Extraction of clauses, amounts, dates and counterparties from contracts, to bring the data into business systems.
Price lists
Supplier PDF price lists turned into structured tables: codes, descriptions and prices ready to import.
Explore the platform's components
Email document acquisition
How the Mail service, installed at your premises, brings documents arriving in business mailboxes into the flow.
Learn more →IntegrationERP integration
How validated data is written into SAP, Zucchetti, TeamSystem and other systems through the AI agent or the REST API.
Learn more →TechnologyThe multi-engine AI technology
The architecture in which you choose one AI engine for each document model, with 99.8% accuracy with Claude.
Learn more →Frequently asked questions about the IDP platform
What is Data Alchemy's IDP platform?
It's an Intelligent Document Processing platform that covers the entire document cycle: it captures invoices, delivery notes, orders, contracts and price lists, extracts the data with AI without templates, validates it against the ERP master data and writes it straight into the ERP. A single end-to-end flow that eliminates manual data entry.
How does a document go from email to a posted record in the ERP?
In four automatic stages: capture (the document enters from email, folder or API), extraction (an LLM understands the document and extracts its fields with no templates), validation (data is checked against ERP master data and business rules) and writing (validated data is written into the management system through the AI agent or the REST API and webhooks). With Claude Sonnet or higher, reading takes on average around 12 seconds for a one-page document and around 18 seconds for two pages.
What does multi-engine AI extraction mean?
It means the platform supports several AI engines and for each document model you choose which one to use: 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. For example you can process the invoice model on Claude, the delivery note (DDT) model on GPT and the order model on Data Alchemy AI, or enable several models on the same engine. Each document is processed by a single engine, the one paired with its model: engines do not work in parallel on the same document. The stated 99.8% accuracy is measured with Claude; no accuracy figure is published for the other engines.
Does the platform integrate with our ERP or CRM?
Yes. The platform connects to SAP, Zucchetti, TeamSystem, Microsoft Dynamics 365, Oracle NetSuite and AS400 (IBM i) through the AI agent, installed at your premises, which reads master data from the ERP and posts the extracted data into it via a database, CSV/JSON/TXT files or APIs; alternatively, you can integrate the REST API and HMAC-signed webhooks. On the CRM side, the Enterprise licence includes native Salesforce integration. A document extraction API is also available to embed extraction into custom applications and flows.
Do I need to configure a template for each supplier?
No. The platform extracts data without templates: the AI understands the document and finds the right fields even on layouts it has never seen, so there's no need to create or maintain rules for each supplier or invoice format.
What are external fields and remote validation?
They are two capabilities included in the Business and Enterprise licences that verify AI-read data against external sources. With external fields, 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 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.
Which licences are available?
Three, plus a free trial. The Demo is a trial Basic licence: 10 days, 10 documents and 1 model chosen from the predefined ones, with the storage it needs, in a dedicated environment where your data is never shared; the AI agent is not included. The licences are Basic (1 model, 500 MB of storage with a mandatory 7-day retention, i.e. documents deleted automatically 7 days after upload, and pay-as-you-go credits bought online; AI agent not available), Business (everything in Basic plus the Mail service, external fields and remote validation; files kept for up to 60 days from upload, configurable per model; credits in annual quantities; the AI agent as a separately quoted add-on) and Enterprise (everything in Business plus native Salesforce integration, Order Dispatcher and the AI agent included; credits in annual quantities). On Business and Enterprise, credits have annual expiry, payment terms agreed with you and lower unit prices depending on quantity; up to 20% of unused credits carry over to the following year if you renew with the same or a higher volume. Credits are always counted the same way: 1 credit = 1 document, regardless of the number of pages. The comparison is on the pricing and licences page; quotes are on request.
See the platform at work on your documents
With the free Demo you try AI reading on 10 documents for 10 days, with 1 model chosen from the predefined ones, in a dedicated environment. To see the whole flow on your own case — validation against master data and writing into your management system — contact us to plan a call with a demo on your own documents.