What is "Totsugo" (突合)?

Totsugo (突合, pronounced "totsugou") is a Japanese term meaning reconciliation — the process of matching invoices with delivery notes or purchase orders to verify accuracy. Traditionally done manually with Excel, it can now be automated with AI tools like Totsugo.

Can Excel VLOOKUP handle reconciliation?

Technically yes, but formatting variations (half-width/full-width, abbreviations) cause errors, and maintaining formulas becomes burdensome. AI tools like Totsugo solve these challenges automatically.

What types of documents are supported?

Invoices, purchase orders, receipts, delivery notes — virtually all business documents. AI understands even custom layouts.

What is the reconciliation feature?

Upload two PDFs and AI automatically reconciles them. Currently supports PDF × PDF only. Excel and system integrations are coming soon.

What is the transcription feature?

It automatically transfers order data from COREC to freee as invoices — no manual entry required. Currently supports COREC → freee only. More integrations coming soon.

All communications are encrypted and data is securely managed. We may use data for AI learning to improve service quality.

How long to get started?

Account issued same day. Start using immediately with setup support included.

Invoice Automation: Focus on 'Judgment' Not 'Storage' | Complete Guide

“We went paperless, so why is invoice processing still taking hours?”

If this sounds familiar, you might be confusing “digitization” with “automation.” This article clarifies the crucial difference and explains how to achieve true invoice automation.

What is Invoice Automation?

The Goal

True invoice automation means reducing human effort in the invoice processing workflow:

Metric	Before	After	Target Reduction
Processing time	5 min/invoice	30 sec/invoice	90%
Error rate	2-3%	<0.1%	95%+
Staff required	3 people	1 person	66%
Overtime	10 hrs/month	0 hrs/month	100%

The Three Stages

Stage	Action	Human Effort	Automation Level
Storage	Save invoice file	Low	Level 1
Reading	Extract data	Medium	Level 2
Judgment	Verify correctness	High	Level 3

Most “automation” solutions only address Stages 1 and 2, leaving the most time-consuming Stage 3 untouched.

Two Misconceptions about “Automation”

While many companies promote DX in the back office, there are cases where the meaning of “automation” is misunderstood.

Misconception 1: “Digitization is not Automation”

Scanning paper invoices or saving them in the cloud is just Digitization.

Action	Type	Human Work Reduced?
Scanning to PDF	Digitization	No
Storing in cloud	Digitization	No
Emailing instead of mailing	Digitization	Slightly

If humans still need to visually confirm the PDFs and manually check the amounts, the workload remains the same.

Misconception 2: “OCR is not Automation”

Extracting text with AI-OCR is only the first step of Digitalization.

OCR Task	Automated?	What Remains
Text extraction	Yes	Data verification
Field mapping	Partially	Correction of errors
Accuracy checking	No	”OCR caretaking”

Manually checking if the OCR’s output matches the purchase order is what we call “OCR caretaking”—it still requires your constant attention.

Three Levels of Invoice Automation

Level 1: Storage (Digitization)

Feature	Description	Example
What it does	Stores invoice files digitally	Google Drive, Dropbox
Cost	Low/Free	$0-10/month
Time saved	Filing time only	5 min/week
Human work	95% unchanged	Still verify everything

Level 2: Reading (OCR)

Feature	Description	Example
What it does	Extracts data from invoices	AI-OCR tools
Cost	Medium	$100-500/month
Time saved	Data entry time	30 min/day
Human work	70% unchanged	Still verify OCR output

Level 3: Judgment (AI Matching)

Feature	Description	Example
What it does	Verifies correctness automatically	Totsugo
Cost	Medium	$100-500/month
Time saved	Verification time	3 hours/day
Human work	10% of original	Only review exceptions

Comparison by Automation Level

To truly reduce your workload, you need to clarify which level of automation you are aiming for.

Criteria	Level 1: Excel	Level 2: AI-OCR	Level 3: AI Matching
Main Method	VLOOKUP / MATCH	Scanning + OCR	AI Agent Judgment
Cost	Low (Free)	Mid ($100s/mo)	Mid ($100s/mo)
Consistency Check	Manual (Prep needed)	Manual (Visual)	Automatic (AI)
Variation Handling	Impossible (Error)	Impossible (Manual)	Possible (Context)
UX	Cell Entry	Screen Review	Enter Key Approval
Setup Time	Hours	Days	Minutes
Maintenance	Ongoing	Regular	Minimal

Why Automate “Judgment”?

An accountant’s time is consumed not by data entry, but by “Checking (Judgment)” whether the content is correct.

The Questions That Consume Your Time

Question	Difficulty	Frequency
”Is this unit price different from last month?”	Medium	Every invoice
”Does this item correspond to this quote line?”	High	50% of invoices
”Are ‘ABC Inc.’ and ‘ABC Corporation’ the same?”	Low	20% of invoices
”Why is the total $1 off?”	Low	10% of invoices
”Is this a duplicate invoice?”	Medium	5% of invoices

Why Humans Struggle

Challenge	Impact	Traditional Solution
Fatigue	Errors increase	Take breaks
Variations	VLOOKUP fails	Master data cleanup
Complexity	Slow processing	Expert staff
Volume	Overtime required	Hire more staff

Why AI Excels

Capability	Human	AI
Consistency	Decreases over time	Always 100%
Speed	5 min per invoice	5 seconds per invoice
Variation handling	Requires exact match	Understands context
Availability	8 hours/day	24/7

Detecting these “discrepancies” is where modern AI excels. Humans should only confirm what the AI is uncertain about (exceptions). Building this Management by Exception framework is true automation.

Management by Exception

What It Means

Instead of checking everything, you only review what the AI flags as uncertain.

Traditional	Exception-Based
Review 100 invoices	Review 5 exceptions
3 hours work	15 minutes work
Boring, error-prone	Focused, efficient

How It Works

100 Invoices Uploaded
    ↓
AI Processes All
    ↓
95 Auto-Matched → Done automatically
5 Exceptions → Human review

Exception Types

Type	Example	Human Action
Price variance	$100 vs $105	Verify with vendor
New vendor	First-time supplier	Confirm details
Missing PO	No matching order	Investigate
Duplicate	Same invoice twice	Reject one

The Next-Gen Reconciliation Flow with Totsugo

Totsugo does not make “Storage” its goal, unlike traditional receipt SaaS.

Step 1: Drag & Drop

Analysis starts as soon as you drop the invoice. No manual data entry.

Step 2: AI Matching

Automatically matches with PO data at the line-item level.

Step 3: Fuzzy Match

AI absorbs notation variations and layout differences contextually.

Step 4: High-Speed Processing

Rapidly approve “AI-confirmed” items with just the Enter key.

Frequently Asked Questions

Q. What about already-digitized invoices?

A. Totsugo works with existing PDF files. No re-scanning needed.

Q. How accurate is AI judgment?

A. 95%+ accuracy on standard invoices. Exceptions are flagged for review.

Q. Can it integrate with our accounting software?

A. Yes. Currently supports freee with more integrations coming.

Q. What file formats are supported?

A. PDF, Excel, CSV, and more.

Summary: Storage vs Judgment

Approach	Focus	Result
Level 1	Storage	Paperless, but same work
Level 2	Reading	Less typing, still checking
Level 3	Judgment	True work reduction

A blazingly fast experience using only your keyboard—no mouse clicks needed. Why not experience the automation of “Judgment” beyond mere “Storage” today?

👉 Try AI Judgment with Totsugo