Upload two PDFs, and AI reconciles part numbers, quantities, and amounts line by line to surface every difference. You review the results and hit Enter. A 2-hour visual check is done in 10 minutes.
Sign up with your email ・ 14-day free trial ・ Cancel anytime, pay nothing
Our Clients
NABELL Corporation
Manufacturing (Bellows & Expansion Joints)
* Results at NABELL Corporation
Comparing two documents by eye quietly drains your team.
Comparing POs and invoices side by side — part numbers, quantities, amounts, line by line. 2 hours a day. Every day.
Person #1 spends 2 hours checking. But there's no record of "what was checked," so person #2 starts over from scratch.
"M10 Bolt" vs "M-10 Bolt." Same product, different notation — missed. Human eyes have limits.
The AI first pass is already done when it reaches your screen. All that's left: review the results and hit Enter — 10 minutes.
Two core jobs: reconciliation and transcription. AI drafts everything; you confirm with Enter.
Match two PDFs, line by line
Upload two PDFs — like a PO and an invoice — and AI reconciles every part number, quantity, and amount, highlighting only the differences in red. Results export as a report.
"M10 Bolt" vs "M-10 Bolt", "Stainless" vs "SUS304". If they mean the same thing, AI treats them as a match — catching what literal comparison misses.
Rotated scans are corrected automatically. Pages with the same PO number are paired by AI. 200 pages take minutes.
Only look at what differs. Beams visualize exactly what was compared, and every check is recorded.
From documents to data, zero typing
From PDF checksheets into Excel, from COREC orders into freee invoices. AI drafts the transcription; you review and hit Enter.
AI reads numbers and text from PDFs and writes only the values into your own Excel template. Merged cells, formulas, and print settings stay untouched — the layout never breaks. Review and fix cell by cell, then download in bulk.
Auto-generate freee invoice drafts from COREC order data. Review, hit Enter, and the invoice is created in freee.
AI never finalizes on its own. Every AI-written cell is visible at a glance in a confidence-scored draft, approved or fixed by a human — keyboard only. Same steps, same result, whoever runs it.
Rotated PDFs? Auto-corrected. 200 pages matched by PO number? Auto-grouped. Then items reconciled one by one. Every comparison is visualized with beams showing what was matched.
With human double-checks, there's no record of "what was checked and how." So the second checker starts over. With AI, every comparison is traced — people only confirm the diffs with Enter. 2 people × 4 hours → 1 person × 10 minutes.
Manual checks, Excel formulas, RPA, general-purpose AI — an honest look at where each one breaks down.
| Criteria | Manual / Visual | Excel formulas | RPA | General AI (ChatGPT etc.) | Totsugo |
|---|---|---|---|---|---|
| Notation variants ("M10 Bolt" vs "M-10 Bolt") | △ Mental translation; misses happen when tired | × Exact match only | × Stops on anything outside the rules | △ Understands, but inconsistent | ◎ AI judges semantic identity |
| High page volumes (200-page scale) | × 2 hours daily; scales only with headcount | △ Requires manual transcription first | ○ Handles volume | × Forgets content mid-document | ◎ Auto-grouping, done in minutes |
| Resilience to format changes | ○ Humans adapt | × Breaks the moment a column shifts | × Scenario rework required | ○ Robust to layout changes | ◎ AI reads structure; zero setup |
| Audit trail (what was compared) | × Nothing recorded; checker #2 starts over | △ Only formula outputs remain | △ Logs exist, reasoning doesn't | × No way to verify the basis | ◎ Every match recorded with evidence |
| Filling Excel templates (layout preserved) | ○ Layout survives, but slow and error-prone | × Can't read PDFs; manual typing anyway | △ OK if rigid; stops on layout changes | × Regenerates the file; merged cells & formulas break | ◎ Writes values into your template; layout intact |
| Setup effort | ◎ Start now (which is why it persists) | ○ Needs someone who writes formulas | × Scenario development & maintenance | ○ Instant, but unfit for this job | ◎ Just upload to the browser |
* Assessments reflect typical usage patterns; your mileage may vary.
Totsugo is not for everyone. Here is an honest guide before you spend any time on it.
You match two documents line by line, routinely
PO × invoice, outbound × return slips, orders × checklists. Any "compare left and right, row by row" job works, regardless of document type.
Hundreds of checks or transcriptions eat your month
The more volume, the bigger the payoff. Nabel automated ~20K order checks a year — 100×100 PDFs daily.
Your documents are messy: FAX, scans, handwriting
Real-world PDFs, no clean-up needed. No ERP or RPA replacement either.
You want fully unattended automation
Totsugo is designed around human final approval. If you need AI to finalize without anyone looking, this isn't it.
Your documents can't be digitized
Totsugo processes PDFs and scans. Paper that stays paper can't be reconciled — a scanning step is required.
You only reconcile a few documents a month
At $259+/month the ROI won't be there — stick with manual checks and come back when volume grows.
Unsure? Use the 14-day free trial with your real documents and decide. The trial costs nothing.
Nabel — a 199-person bellows manufacturer in Mie, Japan with four overseas sites — was reconciling roughly 100 × 100 PDFs every day by hand. Totsugo took over the first pass.
Every order matched by hand. Month-end was 3 days of nothing but transcription and review.
AI flags the diffs. Humans only confirm the exceptions with Enter.
“Totsugo automates the matching and exception detection — order-check time drops while accuracy improves.”
Nabel — bellows manufacturer in Mie, Japan
Manufacturing, construction, leasing, wholesale, trading. Anywhere humans currently scan two stacks of paper side by side, Totsugo fits in.
Quickly verify the order → ship → bill chain stays consistent
Match historical slips and rein in return tracking
Cut hours from procurement, customs, and AP review
AI handles the matching and difference detection. People focus on the exceptions.
Invoices and POs, orders and checklists. Just upload the two documents you want to compare.
Corrects PDF orientation, groups pages by PO number, and reconciles part numbers, quantities, and amounts. 200 pages in minutes.
Match? Enter. Mismatch? Fix and Enter. What took 2 hours now takes 10 minutes.
Comparing orders with checklists took 2 hours daily. "M10 Bolt" vs "M-10 Bolt" — same or different? Had to check by eye. Now AI does it first, and I just hit Enter.
Matching shipment slips with return slips was painful. Finding the right shipment history across 200 slips. Now AI does it automatically — work time cut to 1/10.
Invoice vs PO amount checks. Hundreds every month, and a ¥100 oversight is a big deal. Now AI checks everything first, so I can just hit Enter with confidence.
Every plan includes a 14-day free trial. Verify the results with your own documents before committing.
Entry plan for small teams to automate document checks
Standard plan for SMBs to run document checks at scale
Full custom plan for high-volume, cross-department processing with dedicated support
* We'll help you pick the right plan for your document mix and volume. Questions before the trial are welcome.
No sales call required. Start at your own pace.
No password needed. Click the magic link in your inbox and your account is ready.
Choose Starter / Business / Enterprise. After registering a card, nothing is charged for 14 days.
Drop in two PDFs and AI starts reconciling immediately. No initial setup, no integrations required.
Cancel during the trial and you pay nothing. For high volumes or custom requirements, talk to us anytime.
Totsugo — the AI cloud for document reconciliation — is live. Learn more about the customer story (2-hour order checks down to 10 minutes), feature set, and pricing in our press release.
CEO, Leach Inc. & Full-Stack Engineer
Graduated from National Kumamoto Technical College Advanced Course. Completed Master's program at NAIST (Nara Institute of Science and Technology).
Early career at 3 startups (Akatsuki, Speee, eXmotion) as engineer intern — all later went public. At Toshiba Software Technology Center, worked on R&D spanning web apps to low-level firmware and embedded systems. Later seconded to Toshiba Tec HQ CDO Office (executive office). Convinced of generative AI's disruptive potential, left to fully commit and founded Leach Inc.
""Build the foundation that keeps operations running smoothly. Protect people from exhausting work through better systems.""
Leach Inc.
CEO Takuya Tominaga
Fudanotsuji Square 9F, 5-36-4 Shiba, Minato-ku, Tokyo 108-0014, Japan
* 14-day free trial on every plan. Cancel during the trial and you pay nothing.
Prefer to talk first? Contact us