Email automationEpicor Kinetic Agent view

How to automate email into Epicor Kinetic

Short answer

Read shared mailboxes through Microsoft Graph or IMAP, classify and extract each message and attachment, match it to Kinetic master data, and create the quote, sales order or receipt through the REST API. Low-confidence items go to a human review queue instead of being posted.

A large share of Kinetic order entry still begins in an inbox: customer POs as PDFs, RFQs in the email body, supplier acknowledgements, and packing lists. A person reads each, finds the customer and part numbers, and keys them in. The work is repetitive and the errors are predictable, such as the wrong ship-to or a revision mismatch.

Epicor offers its own capture and EDI-type options in different editions, and for trading partners that send clean EDI, that is the better answer. Email automation fits the long tail: smaller customers and suppliers that will never adopt EDI but send reasonably consistent documents.

The workable design is not 'AI enters orders'. It is a pipeline that proposes a complete, validated order for a clerk to approve, and gradually auto-posts only the narrow cases where confidence and history justify it.

What teams use it for

Customer PO to sales order

Extract the customer PO number, lines, quantities and requested dates, then create a draft order with matched part numbers and the right price list.

RFQ to quote

Parse requested parts and quantities from the email, look up historical quotes and prepare a quote draft for the estimator.

Supplier acknowledgement to PO release

Match 'we confirm delivery on' emails to PO releases and propose a promise-date update.

Invoice intake

Route supplier invoices to AP with PO match against receipts before posting an AP invoice.

Change and cancel requests

Detect that an email is a change to an existing PO rather than a new order and attach it to the right order for review.

Shipping documents

Capture ASN and packing list emails to pre-fill a receipt against the PO.

Epicor Kinetic integration surface

  • Kinetic REST API (v1 services such as Erp.BO.QuoteSvc, SalesOrderSvc, POSvc, APInvoiceSvc, and v2 endpoints); exact names and methods vary by version, so confirm in the API help for your instance.
  • Customer cross-reference data: customer part numbers and part/customer cross-references to map the customer's part name to your part.
  • BAQs for matching lookups such as open POs by supplier and part, so matching does not depend on table structure.
  • Attachments: Kinetic supports document attachments on many entities, letting you store the original email or PDF with the created record.
  • Epicor's EDI and capture products where licensed, as the better path for high-volume partners.

How it works

  1. 1

    Connect the mailbox

    Use Microsoft Graph with application permissions limited to specific shared mailboxes, or IMAP for other providers. Process by message id to stay idempotent.

  2. 2

    Classify

    Decide message type: PO, RFQ, invoice, acknowledgement, change request, or noise. Start with rules by sender and subject, add a model when rules plateau.

  3. 3

    Extract

    Pull header and line fields from the body and PDFs. Use layout-aware OCR for scans and prefer native PDF text when present.

  4. 4

    Match to master data

    Resolve customer from sender and PO, then each line through cross-references and fuzzy matching. Flag unknown parts rather than guessing.

  5. 5

    Validate business rules

    Check price against price lists, lead time against requested date, credit hold, minimum order and duplicate customer PO numbers.

  6. 6

    Review queue

    Show the clerk the email beside the proposed order with per-field confidence. Edits teach the matcher.

  7. 7

    Post and attach

    Create the record by REST, attach the source file, and reply to the sender or log the order number in the thread.

Example flows

PDF customer PO

Trigger: An email with a PO PDF arrives in the orders mailbox from a known customer.

  1. The pipeline classifies it as a PO and extracts header and lines.
  2. Customer part numbers are mapped through cross-references; two lines match, one is unknown.
  3. A draft sales order is prepared with the unknown line flagged.
  4. The clerk fixes the line, approves, and the order is created with the PDF attached.

Result: Entry time drops from minutes of keying to a short review; errors surface before posting.

Supplier date change

Trigger: A supplier replies that two PO lines will ship later than promised.

  1. The email is matched to the PO via PO number in the subject or body.
  2. New dates are extracted and compared with current promise dates.
  3. A buyer queue item shows old vs new date and affected jobs.
  4. On approval the PO release date is updated and a note recorded.

Result: Buyers see supply risk early and decisions are recorded.

Keeping data on-prem

Email content contains customer pricing and drawings, so many aerospace, defense and electronics shops cannot send it to a public AI service. Keep the whole pipeline inside your network: read mail from Exchange on-prem or Graph through a controlled connection, run OCR and a locally served open-weight model for classification and extraction, and call Kinetic internally. Only the clerk interface and the Kinetic API need to be reachable internally. Log model versions and keep source emails for traceability.

Security checklist

  • Limit mailbox access to named shared mailboxes with least-privilege app permissions.
  • Treat attachments as untrusted: scan for malware and never execute embedded content.
  • Beware prompt injection in email text; the model must not be able to trigger writes, only produce fields that code validates.
  • Keep the original email attached to the record for audit and dispute handling.
  • Separate duties: the person who approves a posting should be accountable under normal Kinetic security.

Ways to get it done

Kinetic EDI or capture products, or an established EDI provider

Effort: Weeks per partner, license cost depends on edition.

High-volume partners with structured documents.

Build in-house with Graph, OCR and REST

Effort: Typically a few weeks for one document type, more for several.

IT and an integrator who can maintain extraction rules and a review UI.

Partner such as Netray with a local-model pipeline

Effort: Scoped per document type; first flow often within a few weeks.

Mixed document types, sensitive data, need for review queue and learning from corrections.

Pitfalls

  • !Customers reuse PO numbers across revisions; define duplicate and revision handling before go-live.
  • !Auto-posting too early creates bad orders that are costly to unwind. Begin with propose-only.
  • !Part cross-references are usually incomplete. Plan clean-up as part of the project.
  • !Email threads carry quoted old messages; extract only from the newest content or you will read stale numbers.
  • !Reply-all auto confirmations can loop. Add loop guards and suppress autoreplies.

Frequently asked questions

Can Kinetic create sales orders from email by itself?

Not out of the box in a way that parses free-form PDFs. Kinetic exposes APIs to create orders, and Epicor sells EDI and capture options, but turning a customer's PDF email into a draft order needs a pipeline for extraction, matching and review.

How accurate is automated PO extraction?

It varies by document consistency. Repeat senders with stable templates can reach high accuracy; new or scanned formats need more review. Use per-field confidence and measure the correction rate on your own mail before relaxing review.

Should the system reply to senders automatically?

Start with a confirmation that includes the Kinetic order number only after a human approves. Auto-replies with wrong data create disputes, and reply loops are a known hazard with automated mailboxes.

Can this run without sending email to the cloud?

Yes. Run classification and extraction on local hardware with open-weight models and local OCR, and call Kinetic inside your network. This is usually required when drawings or export-controlled content appear in attachments.

Which documents should we automate first?

Pick the highest-volume, most consistent one, usually customer POs from your top ten senders. Prove the review-queue loop there, then expand to supplier acknowledgements and invoices.

Want Email automation working on your Epicor Kinetic?

A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.

Get a free assessment