Every custom shop taking work from distributors does the same thing several times a week: open a PDF, read it, and retype it into whatever system runs the shop. Ten minutes a PO, half of it spent squinting at a scan.
It is unbillable, it scales with volume, and it is where transcription errors that cost real money get introduced.
The four ways shops do it now
Retyping into a spreadsheet
Free, universal, and the reason nobody can answer "what did we make on that job" three months later. Works until two people need it at once or something changes and the sheet and the PDF disagree.
Retyping into QuickBooks as an estimate
Better for the books, wrong shape for the shop. QuickBooks has no job status, no proofs, no approval record, and no place for artwork. And QuickBooks Online Plus — the tier with inventory and purchase orders — went to $140 a month in August 2026.
Leaving it in the inbox
The PDF is the record. Works for one person with a good memory, collapses the moment someone else needs to know what is running.
Retyping into shop management software
Better on everything downstream, but the retyping is unchanged. Most systems in this category still start with a blank form.
Why POs are awkward to automate
Anyone who has tried scripting this hits the same four walls.
Layouts differ by platform. A commonsku decorator PO, a Syncore PO and a corporate customer's PO put the same information in different places under different labels.
The supplier field is you. On a distributor PO the distributor is the buyer and you are the supplier or vendor, with the end client in the ship-to block. Naive extraction grabs the wrong company constantly — or worse, grabs the contact person's name and files a job under an individual instead of the business that owes you money.
Totals hide under different labels. Total, Amount Due, Balance Due, Grand Total. And on a decorator PO the total often excludes blank goods being drop-shipped separately, so the biggest number on the page may not be what you are being paid.
Half of them are images. Scanned or printed-then-rescanned PDFs have no text layer. Nothing can be copied out. Text extraction returns an empty string and any automated total comes through as zero.
What good intake actually looks like
- No new habit required. If it needs a portal login or a special upload step, it will not survive a busy week. Forwarding an email is a habit shops already have.
- Attachments travel with it. Artwork ends up on the job record, not in an inbox thread that gets buried.
- Image-only PDFs work. A tool that silently fails on scans fails on a large share of real POs.
- The customer is the buying company. Not the end client, not the contact person.
- Uncertainty is visible. A low-confidence parse should arrive as a draft to confirm, not a wrong job you discover a week later. Anything claiming to be perfect is hiding its errors.
How it works in fulflo
Every shop gets an inbound email address. Forward a PO to it and it becomes a job with the PO number, line items, quantities, dates and customer already populated, and the attachments on the job record.
You can also drag a PDF straight into the new job form if it arrived some other way, or enter it by hand.
Parses the system is not confident about land as drafts with the uncertain fields flagged, so you confirm rather than hunt. The extraction was built against real distributor POs received by a working custom branding shop, which is why it knows the supplier field is you and looks for the buying company rather than the first name on the page. Image-only scans are handled rather than skipped.
From there it is the normal flow — job status your customer can see in a portal carrying your branding, proofs with an approval record, and an invoice back over Stripe, ACH or check into your own Stripe account, with their PO number on it.
Why nobody else does this
Shop management systems in this category start jobs from a blank form. Design tools start from artwork. Distributor platforms generate outbound POs to suppliers, and their electronic PO integrations run to blank goods suppliers, not to decorators. Nobody is solving intake for the shop on the receiving end, which is the position most decorating businesses actually occupy.
Common questions
Can software read a scanned purchase order?
Yes, but only if it is built for it. A scanned PDF has no text layer, so plain text extraction returns nothing and totals come through as zero. Handling image-only POs requires optical character recognition as a fallback, which many tools skip.
What gets extracted from a PO?
In fulflo: PO number, customer, line items with quantities and pricing, totals, dates including in-hands dates, decoration details and notes, plus any attachments carried on the email.
What happens if the parse is wrong?
Low-confidence parses arrive as drafts with the uncertain fields flagged for review, rather than being saved as finished jobs. You confirm or correct before anything enters your pipeline.
Do I have to use a special email address?
You forward to a fulflo inbound address, which works from any email host — Gmail, Outlook, Zoho, a custom domain. You can also upload a PDF directly or enter the job manually.