Purchase order OCR, the front door of the floor.
A customer PO arrives as a PDF. Bulk reads it, matches the lines to your parts and routings, and once someone approves it you have a production-ready order.
From the customer’s PO to jobs on the floor.
Sales order automation in Bulk starts with the customer’s own file and ends with numbered jobs, all linked on one record, so the order is never typed in twice.
Uploaded
Drop the customer’s PDF or scan into Import Documents and pick the configuration for their layout. Each file becomes its own inbound document.
Read and matched
The header and every mapped line are extracted and matched to your parts, processes and routings. The file stays attached to the document.
Checked and submitted
Someone checks the draft and creates the document, marks each line Ready and submits it in one go.
Approved
When a routing on the document needs sign-off, a second person approves it. The submitter never approves their own.
Released as jobs
Approval creates one job per Ready line under one production order, ready to schedule and run.
Upload a batch. Bulk does the typing.
Order entry automation starts with a file, not a form. Choose the configuration for the customer’s layout, add their PDFs or scans, and Bulk reads each one into an inbound document while you carry on with something else.
One configuration per layout
Only active configurations are offered, each set up for one document layout, and every file becomes one order.
PDF, PNG or JPG
Up to 10 MB a file. PDFs with a text layer are read with their text; scanned PDFs are read page by page, up to 10 pages.
Duplicates flagged, not blocked
Each file’s contents are fingerprinted, so a copy you have already uploaded is marked Duplicate, or Same Session within one batch.
Keeps going in the background
Files upload three at a time and carry on while you move around Bulk. Every batch stays in Import History.
Teach it a customer’s layout once.
Document data extraction in Bulk is set up per layout, not per file. An admin uploads one sample PDF, keeps the fields that matter and maps them to your columns, and every later upload of that layout follows the same rules.
Five steps, saved as you go
Upload Sample, Extracted Data, Map Document Fields, Map Line Items and Preview. Leave halfway and pick up where you stopped.
One field sets the line count
Mark a repeating field, such as the item number column, as the driver and Bulk creates one line for each value.
Values cleaned on the way in
Each mapping can split, trim, strip a prefix, parse a date or number, look up your master data, calculate or shift a date before the value lands.
Every save is a version
Version History keeps each saved setup. Rolling back restores an earlier one and records the rollback as a new version.
Every line matched to your parts.
AI order entry only helps if the lines land on your own part numbers. Bulk matches each line to your master data, keeps the value it read, and a line that does not match cannot be marked ready.
Exact first
Part, process and routing codes match your master data exactly, ignoring case. A mapping can switch to fuzzy matching with a threshold, which suits customer names.
The original stays
The value as read, the record it matched, how it matched and the confidence are kept on the line, and the extracted lines themselves are never edited.
Missing means not ready
A line without a part, process, routing or quantity shows Not ready, and the grid names the column to fill.
Customer from the page
The customer is matched from the document, or taken from the configuration, before the document is created.
Check the order like a spreadsheet.
Sales order processing happens on one screen: the document’s fields, its source file, the services you charge and every line in a sheet that saves as you type.
Keys you already know
Enter moves down, F2 or typing opens a picker, and pasted columns and the fill handle work as they do in a spreadsheet. Each cell saves on its own.
Many lines at once
Select rows to mark them ready, apply a service or copy one row’s values onto the rest.
Services priced once
Add deburring, testing or finishing to the document once, with fixed or percentage billing lines, then assign it to the lines it covers.
Paperwork travels with it
The uploaded file is attached to the document. Required forms must be complete before submission, and they follow onto the order or its jobs.
Submitted by one person, approved by another.
Submitting checks the whole document in one pass. If a routing on it needs sign-off, the document locks and waits for an approver who is not the person who submitted it, then the jobs are built in the background.
Sign-off follows the routing
Only routings set to require approval raise a request, one per routing, and the submitter is never among the approvers.
Locked while it waits
A document awaiting approval is read-only for everyone until it is approved, sent back or withdrawn.
Sent back, not started again
Request changes needs a reason and returns the document to In Progress. The submitter can also withdraw it, fix it and resubmit, with nothing lost.
Approved, then built
Once every request is approved, Bulk creates one job per Ready line under one production order. If the run stops, Retry job creation carries on without repeating jobs.
Manual review, or jobs straight away.
Automated order entry is a setting on each configuration. On Manual review people check and submit; on Auto-Create Jobs, Bulk creates jobs as soon as a document is read, for every line that passes the checks you require.
Manual by default
New configurations start on Manual review, so every document waits for a person to check and submit it.
Four checks always on
Auto-create needs a matched part and process, a routing and a quantity on every line. A due date, unit price or total price can be added.
Failures wait for people
Lines that fail a check stay on the document with the missing fields listed, and the file shows Blocked at Production.
A model per layout
Choose Lightning, Fast or Pro for each configuration. Pro, for complex documents, is the default.
When a file will not read, you see why.
Scans are not always clean. Every file in a batch shows where it stopped and what to do next, and a retry picks up from the last saved step instead of starting over.
Retry what can be retried
Failed reading and failed document creation retry from the file, or all at once with Resume Processing. A file rejected at validation is uploaded again.
Automatic retries first
Temporary failures retry up to five times with a growing wait, and after repeated failures reading pauses for a minute before trying again.
Blocked has a reason
On Auto-Create Jobs, a Blocked production step means some lines could not become jobs; open it to see which lines and why.
The batch at a glance
The Overview counts the files complete and how many reached each stage, Upload, Inbound and Production, and says how many failed.
What purchase order OCR should get right.
Purchase order OCR reads a customer’s order document and turns its header and lines into structured data. In Bulk that data becomes an inbound document matched to your own master data, and each answer below has a place in the app where you can check it.
| The question | How Bulk handles it | Where you read it |
|---|---|---|
| Which files can it read? | PDF, PNG and JPG up to 10 MB each. Scanned PDFs are read page by page, up to 10 pages. | Import Documents, Upload Files |
| How does it learn a customer’s layout? | A configuration per layout, set up once from a sample PDF in five steps. | Settings, Inbound OCR |
| Which fields does it capture? | PO number, customer, dates, title and notes, then part, process, routing, quantity, prices, dates and notes per line, plus your custom fields. | Map Document Fields and Map Line Items |
| How are lines matched to my parts? | Exact match on your part, process and routing codes, or fuzzy with a threshold where set; the method and confidence are stored. | Part, Process and Routing columns |
| What if a line does not match? | It cannot be marked Ready until the missing value is filled. | Document items, Status column |
| Who releases the work? | The person who submits, plus an approver when a routing requires sign-off; or Auto-Create Jobs for lines that pass every check. | Approval band; routing settings |
| What if the same PO arrives twice? | The file’s contents are fingerprinted and a copy is marked Duplicate, not blocked. | Upload Files, Selected Files |
| What happens when reading fails? | The file shows Needs attention; extraction and document creation can be retried, and validation failures re-uploaded. | Import History, Upload Session |
Read once, used through to the invoice.
The approved document is where the production order, its jobs, their dates and their prices come from, which is how Bulk runs order to invoice on one record.
- ProductionEach Ready line becomes a job under one production order.
- RoutingA routing needing sign-off sends the document for approval.
- SchedulingJobs arrive with the required date read from the PO.
- InvoicingLine prices and added services travel on to the invoice.
- TeamboardsA KPI can count documents or lines received each day.
Questions, answered.
What is purchase order OCR in Bulk?
It is how Bulk reads a customer’s purchase order and turns it into work. You upload the PDF or scan, Bulk extracts the header and every mapped line using a configuration set up for that customer’s layout, matches the lines to your parts, processes and routings, and builds an inbound document. Once the lines are checked and approved, each one becomes a production job.
Is purchase order OCR the same as sales order automation?
Purchase order OCR is the reading step: Bulk extracts the header and lines from the customer’s PO. Sales order automation is the whole path from that document to an approved order and its jobs, including matching to your parts, the Ready checks and approval.
Which files can Bulk read?
PDF, PNG and JPG files up to 10 MB each. A PDF with a text layer is read with that text, and a scanned PDF is read page by page as images, up to 10 pages. Sharp, straight scans read far better than blurred or tilted ones.
Do we have to set up every customer’s layout?
Once per layout. An admin creates a configuration under Settings, Inbound OCR, uploads a sample PDF and works through Upload Sample, Extracted Data, Map Document Fields, Map Line Items and Preview. Each step saves as you go, and saved setups are kept in Version History so you can roll back.
How does Bulk match PO lines to our part numbers?
Part, process and routing codes are matched exactly against your master data by default, ignoring case. A mapped field can use fuzzy matching with a confidence threshold instead, which suits customer names. The text as read, the record it matched, the method and the confidence are stored on the line.
What happens to a line Bulk cannot match?
It stays on the document but cannot be marked Ready. The grid shows which column is missing, such as Part, Process, Routing or Quantity, and you pick the right record or type the value before marking it Ready.
Does the AI create jobs without anyone checking?
Only if you choose that. Configurations start on Manual review, where people check the lines and submit. Set a configuration to Auto-Create Jobs and Bulk creates jobs straight after extraction for lines whose part, process, routing and quantity all resolved, plus any due date or price checks you add. Lines that fail stay on the document for review.
Who approves an order before it reaches the floor?
Submitting checks the whole document: customer, currency, at least one line, every line Ready with an approved process, and required forms complete. If a routing on the document requires sign-off, it waits for an approver set on that routing, and the person who submitted it cannot approve it. Otherwise it is approved on submit and the jobs are created.
What if the same PO is uploaded twice?
Bulk fingerprints each file’s contents and marks a match as Duplicate, or Same Session when the copy is in the same upload. Duplicates are flagged, not blocked, and a fresh scan of the same paper may not be recognised.
What happens when a file cannot be read?
The file’s Inbound step shows Needs attention with the reason. Failed extraction and failed document creation can be retried from the file, or for the whole batch with Resume Processing, and processing carries on from the last saved step. A file rejected at validation, such as an unsupported type, has to be uploaded again.
Can orders still be entered by hand?
Yes. New Document lets you type the header and lines for a phone order or a walk-in, and it then follows the same review, approval and job creation as an imported document.
Can Bulk replace my ERP?
It can. The approved purchase order becomes the production order and its jobs, carrying their prices and dates, and the same record runs on through stock, invoices and credit notes, installed one app at a time. Where you keep an accounts package or ERP, Bulk syncs both ways with it.
Try it on your worst purchase order.
Start a free trial, set up extraction against your real customer documents and see the order it produces. Then you decide. No card needed.
