Fusion5

Thousands of invoices a month, processed without the data entry

Why New Zealand software provider Fusion5 built Xtracta’s AI data capture into ApprovalPlus — its procure-to-pay solution — to cut cost, tighten spend control and speed up month-end.

“Xtracta has an inherently better OCR capability and a better learning process.”

Bevan Wright

ApprovalPlus Solution Manager, Fusion5

Automation and
control, built into 
the product

Automating repetitive work frees people for higher-value tasks. Tight control over spending heads off nasty month-end surprises. Done well, both flow straight to the bottom line.

That’s the thinking behind ApprovalPlus, Fusion5’s procure-to-pay solution. It takes in the documents that drive accounts payable — invoices, purchase orders, credit card statements and expenses — codes them, runs them through approval, and passes them to the finance system for payment.

But the whole workflow depends on getting data off the document first. With each customer pushing three to four thousand invoices a month, keying that by hand was never going to be a cost-effective use of people. Fusion5 needed data capture good enough to disappear into the background.

Capture, learn,
match, pay

Four years ago Fusion5 switched to Xtracta and wired it into ApprovalPlus through the API; pushing and pulling data between the two systems, no templates required.


Invoices arrive by email

Suppliers’ invoices are forwarded to an Xtracta address. In most cases Xtracta grabs every field it needs and passes the invoice to ApprovalPlus without anyone touching it.


The AI reads any layout

No templates to maintain. Xtracta self-learns each supplier’s design and matches fields against its built-in library of terms — a genuine set-and-forget engine.


Teach by example when needed

If something like an invoice number is missing, an AP user drags a box around it on the scanned image. Xtracta learns from that one example and catches it next time.


Match, approve and pay

ApprovalPlus codes the invoice or matches it to a purchase order, isolates unexpected charges for re-approval, then routes everything through to the finance system for payment.


A solution that pays for
itself,and proves itself

Removing data entry was only the start. The real win is the control and confidence the combined solution gives finance teams — and the way it’s lifted ApprovalPlus itself.

Cost cut to a fraction


Ongoing charges run at roughly a tenth of the old OCR approach, with data-entry labour taken out of the process entirely.

Control over every dollar


Customers now enforce purchase orders and set procurement parameters — fewer surprises, far less spend slipping through unchecked.

Audit-ready, faster close


Every invoice carries a stamped audit trail and month-end speeds up — so finance can report sooner, with confidence.

Read more Customer Stories

  • Global customs brokerage using AI data extraction for shipment documentation

    Global Customs Brokerage

    For a global customs brokerage processing high-volume shipments across Canada and the United States, legacy “fingerprinted” templates created manual bottlenecks whenever invoice formats changed. Xtracta replaced rigid OCR with intelligent data extraction, helping the team process complex customs documents faster, more consistently, and at scale.

    Read the case study arrow pointing to the right: Global Customs Brokerage
  • BestStart Education

    BestStart was paying to outsource invoice receipting. They wanted it back in-house but didn’t want to add manual work. Xtracta plugged into their accounting system and now processes over 4,000 invoices a month. More accurate than the outsourced service, and it keeps getting better the longer it runs.

    Read the case study arrow pointing to the right: BestStart Education
  • Prospend

    Prospend needed document capture that could handle invoices and receipts for a global customer base. They chose Xtracta for the technology, the pricing, and the hands-on support. The white-label platform went live quickly, branded as Prospend’s own, and has been running reliably ever since.

    Read the case study arrow pointing to the right: Prospend