From scanned PO to QB Estimate in 45 seconds
Today's PO landed in the Gmail inbox at 1:39 AM Pacific. A scanned image, forwarded by a customer called Datamoto. Two line items: 10 Rollerblades at $123 each, 5 Gas Can Sunglasses at $90. Total $1,680. I ran it through SideQuest at 8:54 AM.
0:00 Asked Claude: list my incoming purchase orders and draft an estimate for the first one.
0:03 Claude pulled the inbox label, found the Datamoto email, parsed the PNG attachment. Returned: PO-01103, two lines, customer Datamoto.
0:08 Catalog match attempt. Both line SKUs (DM19012 and DM78123) returned unmatched against the QB sandbox catalog. Confidence 0.34 and 0.43. Claude flagged both for review.
0:12 Claude tried to find Datamoto as a customer in QB. Not there. Flagged for review.
0:15 Built a local draft with both lines flagged "needs SKU" and customer field blank. Attempted to submit to QuickBooks. QB API rejected because the customer reference was null. Draft stayed in local-review state. No half-baked record made it into QB.
0:20 Asked me how I wanted to resolve. Offered three options: map both lines to an existing "Services" item (fastest), create real items in QB for both SKUs (cleaner), or hold for manual review. I picked the middle option.
0:25 Claude drove the QB web UI via the browser MCP. Created two new Products and Services entries: DM19012 Rollerblade ($123) and DM78123 Gas Can Sunglasses ($90).
0:32 Created a new QB customer called Datamoto. Captured the customer ID (58) from the resulting URL.
0:38 Discarded the original draft. Re-built it with the new customer ID and new SKU mappings.
0:42 Submitted the new draft to QuickBooks via the API. QB accepted. Returned Estimate ID 145, doc number 1001.
0:45 Opened Estimate 1001 in the QB UI to verify. Subtotal $1,680.00. Customer: Datamoto. Both lines correct.
What happened in 45 seconds
The work that got done: parse a scanned PO, create two catalog items, create one customer, build and submit an Estimate. The work I did: type one sentence and approve one decision.
Most POs don't require new customers or new catalog items. The same PO from Datamoto next week would resolve in 8 seconds, because Datamoto and both SKUs are now in QB and the cross-references are seeded.
If you're a distributor running QuickBooks and burning 4 minutes per PO on data entry, this is the loop SideQuest gives you. Most of the value is the next 12,000 POs after this one.
Keep reading
The five PO formats that break every OCR pipeline
We ran 270 real industrial-distributor POs through three off-the-shelf OCR services. About 30% came back unusable. The five shapes that cause it, and …
TechnicalReading handwritten POs without losing your mind
The handwritten PO is the test case nobody wants to talk about. Phone photos, faxes, pen-corrected scans. Here's how the vision pipeline handles them,…
EngHow we evaluated 12 OCR services for SideQuest (and which one won)
Per-page price, accuracy on bordered tables, real numbers. Azure Document Intelligence primary at $0.01/page, Claude Sonnet 4.6 vision as a rescue pas…