2,458purchase orders arrived in 30 days

0purchase orders typed in by hand

4 minmedian from email to ready to confirm

~156 hof hands-on time handed back each month

The challenge: every emailed order retyped by hand

Clarus Adventure runs on NetSuite, and its customers send purchase orders the way most customers do: by email, as PDFs, CSVs and spreadsheets in whatever layout their own systems produce. Every one of those orders had to be opened, read and keyed into NetSuite as a sales order. That’s 5–10 minutes of hands-on time per order, multiplied across more than two thousand orders a month.

Retyping at that volume is slow, it’s easy to get a quantity or item code wrong, and it ties up people who could be looking after customers instead.

The solution: APEX Document Processor

Clarus Adventure uses APEX Document Processor to take purchase orders from inbox to NetSuite:

Orders arrive. APEX watches the order inboxes every 5 minutes, around the clock.
AI reads each one. APEX’s AI extracts the customer, PO number, order date, delivery address and every line item, and gives each order a confidence score.
Matched to NetSuite. APEX recognises repeat customers, translates customer SKUs and branch codes to Clarus Adventure’s own items, and flags anything uncertain, such as an unmatched item, a price that doesn’t line up or a possible duplicate.
Reviewed and created. The team checks each order beside the original document, and one click creates the sales order in NetSuite, linked back to the source document for a full audit trail.

APEX review screen showing a customer's purchase order on the left and the extracted NetSuite order fields on the right, with customer details blurred. — The review screen: the original purchase order on the left, the extracted order on the right. Customer details are blurred.

The results

In a 30-day period, 2,458 purchase orders arrived and none was typed in by hand. APEX created 2,381 sales orders in NetSuite with zero errors.

Measure	Result	Based on
AI extraction time	4.2 seconds (median), 90% under 10 seconds	2,747 orders over 90 days
Email received to ready to confirm	4 minutes (median)	6,158 orders over 90 days
Orders timed through to a NetSuite sales order	2,346	Last 30 days
Hands-on time per order	Under 1 minute, down from 5–10 minutes	Review and confirm in APEX
Time handed back	About 156 hours a month	2,346 orders × 4 minutes saved

The 156 hours uses the conservative end of the range: 5 minutes to key an order by hand against under a minute to review it in APEX. At 10 minutes per manual order, the saving is more than double that.

Confidence that improves with review

Every order gets a confidence score when it lands. Across the period, 51% of orders landed at 85% confidence or higher. Average confidence was 71% when orders landed and 80% when they were submitted to NetSuite, and on average a reviewer’s corrections added 7 points to an order’s score. That’s the point of the design: high-confidence orders move straight through, and the uncertain ones get a person.

Confidence Over Time chart tracking AI extraction confidence scores across the period. — AI extraction confidence over the period.

Daily Volume chart showing purchase orders received, completed and errors per day over roughly one month. — Daily volume: orders received, completed and errors per day.

On every order, the same checks

Repeat customers are recognised and matched instantly.
Customer SKUs and branch codes are translated to Clarus Adventure’s own items.
Duplicate orders are blocked by two layers of checks.
Stock and warehouse restrictions are flagged up front.
Nothing reaches NetSuite without a person’s sign-off.

See it for yourself

Watch the 85-second product film, which uses real screens from Clarus Adventure’s order queue. To estimate what the same automation would save your team, try the ROI calculator.