Invoice Capture
Invoice capture is the first step of accounts payable: receiving supplier invoices by email, upload or scan and turning them into structured data, such as supplier, invoice number, dates, line items, tax and totals, ready for matching and approval.
Last updated
Quick answer
What is invoice capture?
Invoice capture is the first step in accounts payable: collecting supplier invoices from email, uploads, scans or e-invoicing and turning them into structured data such as supplier, invoice number, dates, line items, tax and total. Modern capture uses OCR and AI to extract the fields, and sends uncertain ones to a person to check.
Formula
Field accuracy = Fields captured correctly without correction ÷ Total fields captured × 100
Example: Across 200 invoices with 10 header fields each (2,000 fields), reviewers corrected 60 fields. Field accuracy = 1,940 ÷ 2,000 × 100 = 97%.
Invoice capture turns an invoice into data the AP system can use. Capture that reads line items, not just header totals, is what makes automatic PO matching and touchless processing possible. Ardent Partners reports that 57.4% of suppliers submit invoices electronically and 35.4% of invoices are processed straight-through on average.
Key Industry Data
- 57.4% of suppliers submit invoices electronically; 35.4% of invoices processed straight-through on average: Ardent Partners, State of ePayables 2025 (2025)
Definition
Capture used to mean typing invoice details into the accounting system. Today it is usually automated: OCR reads the text, AI identifies the fields and line items, and anything uncertain is flagged for review. Good capture handles PDFs, scans, photos and e-invoices, reads line items as well as header totals, and checks for duplicates as invoices arrive. Everything after capture, from matching to approval to payment, depends on the data being right.
How invoice capture works
Capture has four stages. Intake collects the invoice from wherever it arrives: a shared AP mailbox, an upload, a scanner or an e-invoicing network. Recognition turns an image or PDF into text; for a scan or photo this is OCR, while a digital PDF or e-invoice already contains the text. Extraction works out which text is which field: supplier, invoice number, dates, PO number, currency, line items, tax and total. Validation checks the result, for example that the lines add up to the total, the supplier exists in the vendor list, and the invoice number has not been seen before.
Fields the system is unsure about are sent to a person to confirm. That review step is part of capture, not a failure of it. The aim is that people check the few fields in doubt instead of keying every field on every invoice.
Header capture versus line-item capture
Header capture reads the supplier, invoice number, dates and totals. It is enough to record a bill and pay it. Line-item capture also reads each line: description, quantity, unit price and amount. You need line items to match an invoice to a purchase order line by line, to code different lines to different GL accounts, and to spot a price change on one item in a 40-line invoice.
OCR, templates and AI
Early capture tools used OCR plus a template for each supplier layout, which broke whenever a supplier changed its invoice. AI extraction reads the layout the way a person would, so it works across suppliers without a template per vendor. Both still need a review step for low-confidence fields.
| Method | How it works | Best for | Limits |
|---|---|---|---|
| Manual keying | A person types each field into the accounting system | Very low volume | Slow; typing errors flow into payment |
| OCR with templates | OCR reads text; a template per supplier layout says which text is which field | A few suppliers with stable layouts | Breaks when a supplier changes its invoice |
| AI extraction | A model identifies fields and line items on any layout, with a confidence level | Many suppliers and layouts | Low-confidence fields still need review |
| E-invoicing | The invoice arrives as structured data | Suppliers on a network or mandate | Needs suppliers to send e-invoices |
| System | What it offers |
|---|---|
| QuickBooks Online | A bills inbox on every tier: email or upload invoices and AI pre-fills bills with line items, amounts and due dates. |
| Xero | Upload or forward bills and Xero’s AI creates draft bills ready for approval. |
| NetSuite | Bill Capture takes PDF, JPEG or PNG invoices by email or upload and extracts fields with a generative model; email intake needs the Transaction Email Capture SuiteApp. |
| Business Central | An optional OCR service turns PDF invoices into purchase invoices, and the Payables Agent drafts invoices from a monitored mailbox. |
| Sage Intacct | The separate AP Automation subscription creates AI drafts from emailed or uploaded invoices. |
Sourced from each vendor’s own documentation; see our integration pages for details and dates checked.
Why It Matters
Capture is where most AP errors start. A mistyped amount, a wrong invoice number or a missed line flows into matching, approval and payment. Manual keying is also slow, so invoices sit in a mailbox while discount windows and due dates pass. Ardent Partners reports that 57.4% of suppliers submit invoices electronically on average, which leaves a large share still arriving as paper or images that need recognition.
Examples
Email capture
A supplier emails a PDF to the AP inbox. The system reads the supplier, invoice number INV-20931, invoice and due dates, five line items and a $3,416.50 total, confirms the lines add up, and queues the invoice for matching.
Line-level capture
A 40-line distributor invoice is captured line by line. Line 17 shows a unit price of $8.75 against $8.25 on the PO. On 120 units that is a $60 overcharge that a header-only total would have hidden.
Duplicate caught at intake
The same supplier emails invoice 5512 for $1,980 on the 3rd and again with a payment reminder on the 18th. Capture recognizes the second copy as a likely duplicate before anyone codes or approves it.
Common Mistakes
Capturing only header totals
If you match to purchase orders or code lines to different accounts, capture line items. Totals alone cannot show which item changed price.
Several inboxes for invoices
Invoices sent to individual staff get lost and duplicated. Publish one AP address to suppliers and forward everything there.
Skipping review of low-confidence fields
Let the system flag what it is unsure of and have someone confirm those fields before matching. A wrong amount is far cheaper to fix at capture than after payment.
No duplicate check at intake
Compare supplier, invoice number, date and amount against history as each invoice arrives, including near-identical invoice numbers.
Measuring capture speed but not accuracy
Track how many fields are corrected per invoice. Fast capture that needs heavy correction just moves the work.
In Practice
- Set up one AP email address and ask suppliers to send invoices only there.
- Ask your largest suppliers for digital PDFs or e-invoices instead of paper.
- Decide which fields you need: header only, or header plus line items for PO matching.
- Turn on duplicate checks against invoice history at intake.
- Review low-confidence fields daily so invoices do not wait.
- Track field accuracy and the most-corrected fields each month, then fix the causes, such as poor scans or missing PO numbers.
Tools and Further Reading
- Touchless invoice processing: What accurate capture makes possible downstream.
- Straight-through processing: How to measure invoices that need no manual work.
- Cost to process an invoice: Published benchmarks for manual and automated processing.
- Cost per invoice calculator: Work out your own cost per invoice.
- NetSuite Bill Capture explained: What NetSuite’s native capture does and does not do.
- Free accounts payable course: Seven self-paced modules covering the AP process, matching, approvals, paying bills, the close and AP metrics.
How Aequo AP Helps
Aequo AP receives invoices through a dedicated forwarding inbox or by upload. AI extracts the header and line items for review, and duplicate attachments are flagged on the way in. Each invoice is also compared with past invoices from the same vendor, including near-identical invoice numbers and amounts within about 2%, and given a risk level before it is matched, coded and approved.
Start Free TrialFrequently Asked Questions
What is the difference between OCR and AI invoice capture?
OCR converts the image into text. AI capture goes further by working out which text is the supplier, invoice number, dates, totals and line items, even when every supplier uses a different layout.
What data does invoice capture extract?
Usually supplier name and details, invoice number, invoice and due dates, PO number, currency, line items with quantities and prices, tax and totals.
How accurate is automated invoice capture?
It depends on the tool and the quality of the invoice. Good systems show a confidence level and send uncertain fields for review, so errors are caught before matching.
What is the difference between invoice capture and invoice processing?
Capture is the first step: turning the invoice into data. Invoice processing is the whole workflow after it, including matching, coding, approval, posting and payment.
Can invoice capture read handwritten or poor-quality scans?
Partly. Clear digital PDFs give the best results. Faint scans, photos at an angle and handwriting reduce accuracy, so more fields go to review.
Is e-invoicing the same as invoice capture?
No. An e-invoice already arrives as structured data, so it needs little or no recognition. Capture is what turns PDFs, scans and photos into the same kind of data.
Category
coreDeep Dive Guides
ai document
Product capability and workflow fit.
Invoice Capture
Stop typing invoice data. Let AI capture and extract data from any invoice format in seconds.
automated invoice processing
Product capability and workflow fit.
Prevent Duplicates
Catch duplicates before you pay them. AI detects duplicates that manual review misses.
Ready to automate your AP?
See how Aequo AP can transform your accounts payable process.