Can't ChatGPT or Claude do this?

No. General-purpose AI is great at "summarize this PDF," but it can't "reconcile 200 pages of POs with 200 pages of invoices, item by item." Two reasons: (1) Page limit: general AI can't process long documents at once — it forgets along the way. (2) Structure: accurately reading which cell is which item from a PDF layout requires a specialized engine. That's what "M10 Bolt" vs "M-10 Bolt" matching needs. The same goes for transcription: ask a general AI for an Excel file and it regenerates the workbook, breaking merged cells, formulas, and print settings. Totsugo is purpose-built for reconciliation and transcription — it writes only the values into your original template.

Can we verify the accuracy before committing?

Yes. When you get in touch we will walk you through checking the matching accuracy and speed on the documents you already work with. Pricing depends on your scale and the systems you connect, so we quote individually.

What if the AI misses something or gets it wrong?

Every match is displayed with its evidence — exactly which part of the left document was compared with which part of the right. AI never finalizes anything on its own: a human always makes the final call. If you spot a mistake, fix it on the spot and confirm with Enter. There is no "AI silently made a mistake somewhere" failure mode.

What types of documents are supported?

Invoices, purchase orders, receipts, delivery notes — virtually all standard business documents. AI understands even custom layouts. Faxed documents and PDFs with handwriting work as-is.

Does Totsugo also do transcription, not just reconciliation?

Yes. Totsugo transcribes PDF checksheets into your own Excel templates, and auto-creates freee invoice drafts from COREC order data. Excel transcription writes only the values into your original template, so merged cells, formulas, and print settings never break. In both cases AI drafts the work and a human reviews cell by cell, then confirms with Enter.

Do we need to change our workflow?

No. Just upload the PDFs you already use. No format changes or system integrations required.

How is my uploaded data handled?

All traffic is encrypted via SSL/TLS, and stored data is encrypted at rest. Your documents are never shared with third parties without your consent. See our Privacy Policy for details.

How long does setup take?

Create an account with your email, pick a plan, and you can start immediately. The whole flow is online and self-serve — no initial setup needed.

Invoice Confirmation Matching | Verify Purchase Orders and Deliveries

Trust, but verify.

When an invoice arrives, you need to confirm it matches what was ordered and what was received. This process—confirmation matching—is your defense against overpayment, fraud, and errors.

This guide covers confirmation matching from basics to automation.

What is Confirmation Matching?

Definition

Confirmation matching is the process of verifying invoice details against supporting documents (purchase orders, delivery notes, contracts) before approving payment.

Think of it as a checkpoint:

Invoice claims X
Documents show Y
Do X and Y match?

Why “Confirmation”?

You’re confirming that:

The goods/services were actually ordered (PO exists)
The goods/services were received (delivery note exists)
The amounts match what was agreed (prices correct)

What You’re Preventing

Risk	How Confirmation Matching Helps
Overpayment	Verify quantities match delivery
Fraud	Confirm PO was authorized
Errors	Catch price/calculation mistakes
Duplicates	Match to specific PO/delivery

The Confirmation Process

Step 1: Receive Invoice

Invoice arrives via:

Email attachment
Vendor portal
Mail (scanned)

Key fields to extract:

Vendor name
Invoice number
Invoice date
PO reference
Line items
Total amount

Step 2: Locate Source Documents

Find the matching:

Document	Where to Find	Key Identifier
Purchase Order	ERP system	PO number on invoice
Delivery Note	Receiving system	Delivery reference
Contract	Contract management	Vendor/project ID

Common Problem: Documents are in different systems.

Step 3: Compare Key Fields

Field	Invoice	PO	Delivery	Status
Quantity	100	100	95	❌ Mismatch
Price	¥1,000	¥1,000	-	✅ Match
Total	¥100,000	¥100,000	-	❌ Should be ¥95,000

Step 4: Investigate Discrepancies

For each mismatch:

Document the variance
Determine if it’s material
Contact vendor or internal team
Resolve before payment

Step 5: Approve or Reject

Decision	When
Approve	All fields match within tolerance
Partial approve	Request credit for overcharges
Reject	Significant discrepancies
Hold	Awaiting clarification

Common Discrepancies

Quantity Mismatches

Scenario	Example	Action
Under-delivery	Received 95 of 100	Pay for 95
Over-delivery	Received 105 of 100	Return 5 or approve extra
Wrong count	Invoice says 100, received 90	Request correction

Price Discrepancies

Scenario	Example	Action
Price increase	PO: ¥1,000, Invoice: ¥1,100	Verify contract terms
Discount missing	10% discount not applied	Request credit
Wrong currency	USD vs JPY confusion	Clarify

Timing Issues

Scenario	Example	Action
Invoice before delivery	Invoice dated 1/1, delivery 1/15	Hold until confirmed
Old PO	Invoice references expired PO	Verify renewal
Duplicate invoice	Same invoice sent twice	Reject duplicate

Tolerance Levels

Not every variance needs investigation.

Setting Thresholds

Variance Type	Common Tolerance
Quantity	1-3%
Unit price	0-2%
Total amount	¥1,000-5,000

Example Policy

Invoice Total	Tolerance
< ¥50,000	¥500 or 3%
¥50,000-500,000	¥2,000 or 1%
> ¥500,000	¥5,000 or 0.5%

When to Investigate

Always investigate if:

New vendor (no history)
First invoice for new contract
Unusual items or services
Above-threshold variance
Manual override requested

Manual vs. Automated Matching

Manual Process Time

Task	Time
Find PO	2-5 min
Find delivery note	2-5 min
Compare line items	5-10 min
Document findings	2-3 min
Total	11-23 min/invoice

Automation Benefits

Factor	Manual	Automated
Time per invoice	15+ min	1-2 min
Error rate	1-3%	<0.5%
Scalability	Limited	Unlimited
Audit trail	Manual	Automatic

AI-Powered Confirmation

How AI Helps

Extract invoice data automatically
Match to PO and delivery records
Highlight discrepancies
Suggest resolution actions
Learn from corrections

The Workflow

Invoice received
      ↓
AI extracts data
      ↓
AI matches to PO/delivery
      ↓
Discrepancies flagged
      ↓
Human reviews exceptions only
      ↓
Approve with Enter

Handling Variations

Challenge	AI Solution
“ABC Corp” vs “ABC Corporation”	Fuzzy matching
Different date formats	Normalization
Line item descriptions vary	Semantic matching
Multi-page invoices	Document understanding

Best Practices

1. Require PO Numbers

Reject invoices without valid PO references.

2. Centralize Documents

Store POs, deliveries, and invoices in accessible systems.

3. Set Clear Policies

Document tolerance levels and escalation paths.

4. Track Problem Vendors

Flag vendors with frequent discrepancies.

5. Automate Routine Matches

Focus human effort on exceptions.

Integration Considerations

With ERP Systems

System	Integration
SAP	Verify against MM/SD
Oracle	Link to PO module
NetSuite	Match via transactions

With Accounting Software

Software	Consideration
freee	Match to registered invoices
QuickBooks	Verify vendor/item records
Xero	Link to purchase orders

Summary

Confirmation Matching Steps

Receive invoice
Locate source documents
Compare key fields
Investigate discrepancies
Approve or reject

Key Controls

Control	Purpose
PO matching	Authorized purchase
Delivery matching	Goods received
Price matching	Correct terms
Tolerance levels	Focus on material variances

Key Takeaways

Confirmation = Trust but verify
3-way matching catches billing errors
Tolerances prevent wasting time on immaterial variances
AI automates routine matching, humans handle exceptions

Stop paying blindly. Confirm before you pay.

👉 Try PDF reconciliation for free