“Honestly, checking work is far more rewarding for the team than typing it in.”
Order capture
Your orders come in by email. They go out to your ERP.
The agent extracts every order (PDF, Excel, plain text) and records it in your ERP. What it reads without ambiguity goes through on its own, the rest comes back to you. 70% less data entry time, from 10 minutes to 3 minutes per order.
Use cases
Two ways an order arrives.
The two situations our customers meet most.
No attachment
The order is in the message
The customer types references and quantities straight into the email. The agent reads them, matches them against your catalogue and prepares the draft.
- 40%through with no hands
- 70%less data entry
PDF or spreadsheet
The order is attached
The purchase order comes as an attachment, sometimes across several pages. The agent reads the header and the lines, fills the ERP, and leaves your teams what the document does not say.
- 10 daysto be running
- 3 h/dayfreed per person
Hello,Please deliver 50 XJ-200 and 12 XJ-204.Delivery to the Roanne warehouse.Best regards, M. Dupont
ReadingWhat the agent read
CaptureFrom: achats@ateliers-berger.fr · BC-2026-0912.pdf
What the agent read
How it works
From email to ERP, no re-typing.
Every order email becomes a structured order in the ERP. What is read without ambiguity goes through on its own, the rest waits for your teams. Data entry disappears. Control stays.
From email to the order recorded in your ERP.
The agent identifies order emails, extracts the line items (PDF, Excel, email body) and creates the order. When something is left to settle, your teams settle it in the interface, integrated into your ERP or CRM. Every order stays traceable from the original email to the record.
- Line extraction from PDF, Excel or plain text, including multi-page attachments
- The order recorded in the ERP, left as a draft when something is missing
- Full traceability: every order tied back to the email it came from
Tags landed
ExtractionRead from the document
The agent captures. It asks when it is unsure.
When the reading is clean, the order reaches the ERP with nobody touching it. When it is not, it stops: an unknown reference, an unreadable quantity, a missing date, and the line waits for one of your team. Every order stays linked to the email it came from.
- What is read without ambiguity goes through untouched
- What is not stops, and is put in front of your team
- Every extraction stays traceable back to the email it came from
Orders and quotesIn progressTo complete3
Sent to the ERP18
Failed1
Order capture writes into the system you already run.

- Infor
- and more
Features
Four things the agent does very well.
PDF · XLS · TXT
Multi-format
PDF, Excel, free text: the agent reads the formats your customers actually use, including page-range detection in PDFs.
On doubt
It stops when unsure
An unknown reference, an unreadable quantity, a missing date: the line stops and comes back to you. The rest goes through.
ERP · CRM
Integrated with your tools
Interface built directly into your ERP or CRM, or in the Catalog web app.
AI
Learning
You correct once, the agent learns. You validate, it applies from then on. A learning never touches your ERP data.
Questions
Frequently asked questions
Do we need to connect our ERP?
Yes, that’s where orders end up. Connection via API, webhook, or flat files: SAP, Sage, Cegid, Microsoft Dynamics, Sellsy, EBP, Divalto, AS/400, and several industry-specific or in-house ERPs. The interface can integrate directly into your ERP or CRM.
What measurable return can we expect?
70% less data entry time: from 10 minutes to 3 minutes per order, measured across our customers.
How long to get it live?
10 days to go live, measured across our deployments.
Where is our data hosted?
Exclusively in the European Union, with no transfer outside the EEA. TLS 1.2+ in transit and AES-256 at rest, databases isolated per customer. Your emails never train AI models: that is a contractual constraint imposed on our providers. A DPA is available for signature.
Which ERPs are supported?
Catalog integrates via API, webhook, or flat files. In production today: SAP, Sage, Cegid, Microsoft Dynamics, Sellsy, EBP, Divalto, AS/400, and several industry-specific or in-house ERPs.
In their words
The diagnostic
See how many orders are waiting to be entered.
We ran the analysis on one company’s mailbox: the ordinary mailbox where their orders and quotes arrive. Here is what sixty days of it held.
- The sample
- From a single analysed mailbox, over sixty days. Volumes are given as a daily rhythm and proportions are rounded, so that no company reads its own counts here. Yours will read differently.
- 48 hours
- From the export landing with us to the report in your hands. A commitment, not an average.
- 90 days
- The window the report is built on. Long enough to cover a quarter, and your slow weeks with it.
- One file
- A mailbox export, which your IT team makes in an afternoon. That is the whole of what we ask.
- No ERP access
- Nothing is connected, nothing is installed, nothing is written. The report reads mailbox data and says so.