Prophet 21 Order Entry: Every Way to Get Customer Orders into P21

Epicor Prophet 21 is built for distribution — part numbers, selling units, ship-tos, price pages and contracts are all first-class. What it does not do is read. Every order that arrives as an email, a PDF purchase order, a spreadsheet or a WhatsApp message still has to become a P21 sales order somehow, and for most P21 distributors “somehow” means a person on the inside sales desk retyping it.

This guide covers every route an order can take into Prophet 21 — manual entry, the scheduled file import, EDI, the API and AI order intake — what each one needs, where each one breaks, and how to decide which of your orders should take which route. It is written for operations and IT leads at P21 distributors, so it stays at the level of decisions rather than field mappings.

What a P21 sales order needs before it will save

Whatever route an order takes, Prophet 21 wants the same things before it will accept it, and most failed imports and API calls come down to one of them being missing or ambiguous:

• A customer that resolves to a real P21 customer record — not a company name typed slightly differently.
• A ship-to. Many P21 customers have several, and an order that does not say which one is not finished.
• Items that resolve to P21 item IDs. The customer's own part number or description has to be translated to yours first.
• A unit of measure P21 recognises for that item — a case is not an each, and a case of 12 is not a case of 24.
• The sales location the order ships from, if you run more than one.
• Pricing that matches what P21 would calculate from that customer's price pages and contracts, or a deliberate reason it does not.
• The customer's PO number, carried through so acknowledgements and invoices match their paperwork.

Keep that list in mind, because it is the real test of every method below. Getting text into Prophet 21 is easy. Getting a sales order that resolves every one of those fields correctly, without someone fixing it afterwards, is the actual job.

Route 1: Manual entry in Order Entry

Still the default for most P21 distributors. Someone reads the order, opens Order Entry, finds the customer and ship-to, looks up each item, picks the unit, keys the quantity and the PO number, checks the price and saves.

Its strength is that a person can read anything — a scrawled order sheet, an email that is half order and half complaint, “same as last week but no tea”. Its cost is that every one of those steps happens in sequence for every order, so order volume and headcount move together. It is also where most order errors start: not in the typing, but in the matching — the right item at the wrong unit, or a customer part number mapped to the nearly-right product under time pressure.

Route 2: Scheduled file import (SISM)

Prophet 21's native bulk route is the Scheduled Import Service Manager, usually called SISM. An external system drops files in an agreed location, SISM picks them up on a schedule, and the orders are created in P21. It is configured by a P21 administrator under the Imports/Exports setup, and orders that fail validation are held for someone to correct rather than committed half-right.

SISM is excellent at what it was designed for: high-volume, template-friendly loads from a system that can produce exactly the layout it expects — a web store, a customer portal, another internal application. What it cannot do is read a file that does not match its template. A customer's own spreadsheet, with their column order and their item descriptions, is not a SISM file until someone reformats it, and reformatting is manual entry wearing a different hat.

So SISM is a delivery mechanism, not a reading mechanism. That distinction matters later: it can be the last step of an automated route, but it cannot be the first.

Route 3: EDI

For trading partners that send EDI, Prophet 21 handles the standard documents: an inbound 850 purchase order becomes a sales order, and the 855 acknowledgement, 856 advance ship notice and 810 invoice go back out. Most P21 distributors run this through an EDI provider rather than maintaining the maps themselves, with rules set up per trading partner.

Once a partner is live, EDI is hard to beat — structured, high volume, little human touch. The limit is reach. Every partner has to be set up on both sides, so EDI tends to cover the large accounts that demand it and stops there. The independent restaurant, the regional chain and the customer who has emailed the same order sheet for ten years will never send you an 850. Our comparison of EDI vs AI order automation goes into where that line usually falls; the short version is that EDI and AI intake cover different customers, not the same customers twice.

One thing EDI does not remove is the exceptions. An 850 with an unknown item or an unexpected unit of measure still lands on the inside sales team. EDI removes typing on clean orders; it does not remove judgement.

Route 4: The Prophet 21 API

Prophet 21 exposes several APIs through its middleware, and they are not interchangeable. Integrators generally describe them like this:

• The Transaction API creates and updates transactions — including sales orders — with P21's business rules applied, which is why it is the usual route for order creation.
• The Interactive API drives P21 the way a user at a screen would. It is the fallback for logic that only exists in the user interface.
• OData is for reading — customers, items, prices and stock — not for creating orders.
• The Entity API covers master records rather than transactions.

Two practical points matter more than the technical detail. First, API access is typically licensed separately, so check what your P21 agreement includes before planning around it. Second, the API has exactly the same demands as Order Entry: if the customer, ship-to, item, unit or sales location does not resolve, the call fails. An API is a door into Prophet 21. It does not tell you what the customer meant.

That is why API integrations work best for orders that are already structured data — an e-commerce order, a portal order, a system-to-system flow. For an order that arrives as a PDF or an email, something still has to turn it into structured data first.

Route 5: AI order intake for everything that arrives unstructured

This is the route for the orders the other four cannot take on their own: email bodies, PDF purchase orders on each customer's own template, spreadsheets in any layout, photos of order sheets and WhatsApp messages. An AI intake layer sits in front of Prophet 21 and does the part a person does today:

• Reads the order — the email body and its attachments together, because an order split between a note and a spreadsheet is still one order.
• Extracts the customer, PO number, items, quantities, units and delivery notes.
• Matches each line to your P21 items and selling units, including the customer's own part numbers and descriptions.
• Checks the result against that customer's pricing and your rules.
• Hands a structured order to Prophet 21, and flags anything it is not confident about for a person instead of guessing.

Notice what that list does not include: a new way into Prophet 21. An intake layer still uses one of P21's own entry routes for the final write. What it adds is everything before that write — turning a customer's message into the fully resolved order that the API or an import requires. That also gives you the cleanest way to evaluate one: ask which route it writes through, and what you see when P21 rejects an order.

Which route for which orders

Most P21 distributors end up using several routes at once, because their customers send orders several ways. A workable split looks like this:

• EDI partners stay on EDI. If it works, leave it.
• System-generated orders — your web store, a portal, another application — go through the API or SISM, because they are already structured.
• Email, PDF, spreadsheet and messaged orders go through AI intake, which prepares them for the same API or import route.
• Genuinely unusual orders — a new customer, a dispute, an order that needs a phone call — stay with a person, wherever they arrived.

To decide where to start, count rather than guess. Pull a normal week of orders and sort them by how they arrived. Outside the largest accounts, the unstructured pile is usually the biggest — and it is the one still being typed by hand.

What about Epicor's own AI?

“AI for Prophet 21” now means several different things, so it is worth separating them. In May 2026 Epicor introduced a wave of agentic AI under its Epicor Prism brand, including a Prophet 21 business rules agent that lets users create and change P21 business rules from natural-language prompts. Reporting on the launch indicates these capabilities are aimed at cloud deployments, so check availability against your own P21 edition.

As announced, those agents are not aimed at reading the purchase orders your customers email you. For unstructured order intake, Epicor's announced route is a partner: in July 2026 it expanded its partnership with Conexiom to bring AI order automation to Prophet 21 and Eclipse distributors. Independent intake tools, ManualOut among them, sit in the same place in the architecture — in front of P21, writing the finished order in.

Whichever you evaluate, the test is the same, and it is not the demo. Give the tool twenty of your own real orders, the messy ones, and check three things: did each line match the right P21 item and unit, did the customer's pricing apply, and when it was unsure, did it say so? There are more questions worth asking in will AI order entry work with my ERP?

What should stay with a person

Automation is only useful if you can trust what reaches Prophet 21. The workable model is human-in-the-loop: orders that are clean and confidently matched are prepared automatically, and anything ambiguous — an item the customer has never ordered, a customer part number with no match, a quantity ten times their usual — is routed to a person with the original message beside it.

That changes what the inside sales desk does all day. Less typing, more deciding: resolving the exceptions, talking to the customer who sent the strange order, catching the account that changed hands. The people stay; the transcription goes.

How to start without disrupting Prophet 21

You do not need to change anything in P21 to find out whether this works for you:

• Pick one channel — usually the shared order inbox — and collect twenty real orders, including the worst ones.
• Check which entry routes your P21 licence includes, API access, SISM or both, because that determines how a finished order gets written.
• Run the sample through an intake layer and compare what it prepares, line by line, with what your desk actually keyed.
• Look hardest at what it flagged. A tool that never flags anything is not more accurate; it is quieter about being wrong.
• Widen to the next channel only when the first one matches.

Where ManualOut fits

ManualOut is an AI order-intake layer that sits in front of Prophet 21. It reads orders from email, PDF, Excel, WhatsApp and file uploads, matches each line to your P21 items and selling units, applies the customer's pricing, assigns the warehouse, flags only the uncertain orders for review and prepares the sales order for P21. The Prophet 21 integration page covers what it reads and prepares; the integration is currently available to qualified pilot customers.

The fastest way to know whether it fits your P21 setup is to watch it run on your own orders. Book a demo and send the messiest email or PDF you have.

Epicor, Prophet 21, Eclipse and Epicor Prism are trademarks of Epicor Software Corporation. ManualOut is an independent product and is not affiliated with or endorsed by Epicor Software Corporation. Prophet 21 capabilities vary by version, deployment and licensed modules; confirm specifics with Epicor or your P21 partner.