# How to automate email for ECI E2 Shop System > Source: https://netray.co/integrations/e2-shop-email-automation > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Integrations ## Short answer Monitor shared mailboxes, classify each message and extract structured fields, match them to E2 customers, parts and orders, then stage the result for a person to approve before it becomes a quote or sales order. Outbound emails use E2's own document output. ## System ECI E2 Shop System ## Capability email-automation ## Intro A typical E2 shop receives customer purchase orders, drawings and RFQs by email all day, to a sales inbox that three people watch. Someone opens the PDF, finds the customer in E2, retypes the lines, and attaches the document. The same happens for vendor order acknowledgments and invoices in a purchasing inbox. The work is repetitive but not trivial. Customers send different layouts, quote numbers are mixed with part numbers, and the same part is called three things in three emails. That is why a plain rule-based parser fails, and why modern extraction models help, provided the output is matched against E2's real data and reviewed. This page treats email as an intake channel into E2: what to read, how to match, what must stay human, and how to keep a clean trail back to the original message. ## Use Cases ### Customer PO to sales order draft **Description:** Read the PO attachment, extract customer PO number, lines, quantities and dates, match to E2 customer and parts, and prepare a draft order for review. ### RFQ intake to quote request **Description:** Detect request-for-quote emails with drawings, create a quote stub with the customer and part numbers, and notify the estimator. ### Vendor acknowledgment matching **Description:** Match supplier order confirmations to open POs, flag changed dates or prices, and show the buyer only the differences. ### Order change requests **Description:** Recognise emails asking to change quantity or ship date on an existing order and attach the request to that order for a CSR to act on. ### Shipping and tracking replies **Description:** Pick up carrier or vendor tracking numbers in replies and add them to the shipment or PO as a note for follow-up. ### Outbound status emails **Description:** Send order acknowledgements and shipment notices generated from E2 documents, using templates and a log of what was sent. ## System Surface SQL Server data for customers, contacts, parts, customer part cross-references, open orders and POs. These form the matching dictionary for extraction. Document output: E2 can print or email standard documents such as quotes, order acknowledgements and POs, depending on version and configuration, which you can use for outbound mail. Import utilities: some E2 versions and add-on modules support importing orders or other transactions from files. Confirm what your licence includes before deciding on a write-back method. Attachments: E2 supports linking documents and drawings to records in several places; the exact mechanism varies. Plan a path to store the original email and PDF with the created record. Mailbox side: Microsoft 365 Graph API or IMAP for shared mailboxes, with application permissions limited to specific mailboxes. ## How It Works ### Connect to the shared mailboxes **Detail:** Use Microsoft Graph with app permissions limited to named mailboxes, or IMAP with an app password. Poll or subscribe to new messages and move processed ones to a folder. ### Classify the message **Detail:** Label each email as PO, RFQ, change request, vendor acknowledgment, invoice or other. Keep an Other bucket for human triage. ### Extract fields **Detail:** Read the body and attachments with PDF text extraction and OCR where needed, then pull out PO number, lines, prices, dates and ship-to. ### Match against E2 **Detail:** Resolve the sender to a customer, map customer part numbers via cross-references and fuzzy match descriptions. Record a match score per field. ### Stage for review **Detail:** Show the original document beside the extracted data with highlighted low-confidence fields. The reviewer corrects and approves. ### Create the transaction **Detail:** Create the order or quote through a supported import or integration, or give the CSR a pre-filled sheet to key in. Link back the email. ## Example Flows ### Customer PO arrives by email **Trigger:** A PDF purchase order arrives in the orders mailbox. **Steps:** - The classifier tags it as a PO and OCR extracts header and lines. - The sender domain resolves to a customer; part numbers map through cross-references. - Two lines have unclear quantity and are flagged. - A CSR fixes them in the review screen and approves. **Result:** A correct draft order is created in minutes, with the original email attached. ### Vendor date change **Trigger:** A supplier emails a revised delivery date. **Steps:** - The email is matched to the open PO by PO number. - The new date is compared with E2 and a difference is flagged. - The buyer sees the change and a suggested impact on dependent jobs. - The buyer updates the promise date in E2. **Result:** Schedule risks surface the day the vendor reports them. ## On Prem If the ERP data and the documents must not leave your premises, run the whole pipeline inside the shop network. The mailbox connector can read Microsoft 365 or on-premises Exchange through the least-privileged account, while text extraction, OCR and the language model run on a local GPU server using open-weight models served with vLLM or Ollama. Matching dictionaries are read from SQL Server with a read-only login, and the review queue lives in a small internal web app. Original emails and extracted data are stored on your own file share or database. The only outbound traffic is the mailbox connection itself. Smaller shops without a GPU can start with a CPU-based OCR and rule hybrid and add a model later. ## Security - Grant mailbox access only to the specific shared mailboxes, using scoped application permissions. - Treat inbound attachments as untrusted: scan for malware and strip active content before processing. - Guard against prompt injection in email text; extraction output is data, never instructions to the system. - Require human approval before any order, quote or price is created, and log reviewer and timestamp. - Validate sender identity, because spoofed POs and changed bank details are common fraud patterns. ## Options ### Rules plus Outlook folders **Fit:** Shops with a few repeat customers with fixed PO templates and low volume. **Effort:** Days, but brittle when layouts change. ### Commercial document intake service **Fit:** Teams who accept cloud processing and need broad support for invoices and POs. **Effort:** A few weeks to configure, plus per-document fees and an E2 hand-off to build. ### Partner build with Netray **Fit:** Shops who need on-prem processing, customer-specific matching and a review UI tied to E2. **Effort:** A scoped project of several weeks, delivered with an E2 domain partner. ## Pitfalls - Auto-creating orders from emails without review leads to duplicated or mispriced orders. - Customer part numbers often differ from internal ones; without cross-references matching accuracy collapses. - Replies and forwards contain old quoted text that confuses extraction; strip quoted history. - Revision letters on drawings are easy to drop; carry revision as its own field. - Ignoring duplicates, the same PO sent twice, creates double orders unless you check customer PO numbers. ## Faq ### Can E2 read emails by itself? **Answer:** Not in the sense of extracting orders from message content. E2 can send documents by email in many setups, but inbound parsing requires an external tool that reads the mailbox and feeds E2 through a supported path or a review step. ### How accurate is automatic PO extraction? **Answer:** It depends on document quality and how well your parts are matched. Clean digital PDFs from repeat customers can reach high accuracy, while scans and unusual layouts need more review. Use confidence thresholds and measure on your own samples. ### Should orders be created automatically? **Answer:** Start with drafts that a person approves. After weeks of data, you may auto-accept only trusted customers with exact matches and a total below a limit. Always keep an audit log and an easy way to reverse. ### Can this work with Gmail or only Microsoft 365? **Answer:** Both work. Microsoft Graph suits Microsoft 365, and IMAP or the Gmail API suits other providers. Use scoped credentials and a dedicated shared mailbox rather than personal accounts. ### What about invoices from vendors? **Answer:** The same pipeline can read invoices, match them to POs and receipts, and stage them for payables review. See the document OCR page for three-way matching, tolerances and coding considerations. ## Search terms e2 shop system email automation, automate purchase orders email e2 shop, e2 shop system rfq email intake, eci e2 order entry automation, email to sales order manufacturing erp, extract po from email pdf erp, e2 shop system vendor acknowledgment, shared inbox automation job shop erp, on-prem email parsing erp ai, e2 shop system quote automation --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json