Accountants at small and mid-sized firms spend an average of two to three hours a day on manual invoice entry. They photograph a paper document, open the accounting software, retype the supplier, invoice number, date, amount, VAT — line by line. Then they do it again for the next invoice.
This work doesn't require accounting expertise. It demands patience and attention to detail. And it eats hours that could go toward real accounting work.
What the invoice-automation AI assistant does
The process is simple: you photograph the invoice with your phone and send it to Telegram. The AI assistant reads the image with Claude Vision, extracts structured data — supplier, invoice number, date, line items, VAT, total — and writes it directly into the import folder of Plaus Minuss. The whole thing takes under 30 seconds.
It works with Bulgarian invoices: Cyrillic, BGN, date formats like "12.05.2026", different VAT rates. No translation or manual intervention is needed for standard documents.
How the accuracy check works
Not every invoice is the same. A photo in poor light, a non-standard supplier font, a handwritten amount in the corner — these cases need attention. That's why the assistant is built with a confidence mechanism:
- High confidence (≥85%): the invoice is uploaded automatically. You receive a Telegram message: ✅ Invoice #1042 from Supplier Ltd — 1,240 BGN. Uploaded.
- Low confidence: the assistant shows you the extracted fields in Telegram and asks whether everything is correct. You confirm or correct — only the fields it's unsure about. Once confirmed, the invoice uploads.
The result: no invoice ever enters the accounting software without human approval when there's doubt. Automation sits where it's reliable. Control stays where it should.
Integration with Plaus Minuss
The assistant generates an XML file in the exact format Plaus Minuss expects and writes it straight into the accounting software's import folder — a local directory, a network share, or SFTP. The next time you open Plaus Minuss, the invoice is already inside.
You don't have to change how you work. Plaus Minuss stays the same. Only the manual entry disappears.
What this means in practice
At 40–60 invoices per month, each handled manually for 3–5 minutes, you're looking at 2–5 hours of lost work time. At higher volumes — significantly more.
Beyond time: transcription errors. A digit slipped in the amount, the wrong VAT percentage, a supplier mixed up — each one demands extra investigation and correction. The AI assistant reads from the original document every time, without fatigue, without careless mistakes.
What it doesn't do
The assistant doesn't replace the accountant. It doesn't post journal entries on its own. It doesn't make decisions about expense categorisation. It doesn't give tax advice.
It does one thing: reads invoices from photos and posts them into the accounting software without manual entry. That task — mechanical, repetitive and error-prone — is exactly where AI works reliably.
What rollout looks like
We start with 20 real invoices from your suppliers to measure accuracy on your specific documents. The process takes four weeks:
- Week 1: process test invoices, measure accuracy by supplier type
- Week 2: configure the adapter for your version of Plaus Minuss and the path to the import folder
- Week 3: live testing with a real accountant in Telegram, set the confidence threshold
- Week 4: parallel run — the assistant and the accountant process the same invoices, compare results
Target before full launch: 95% field accuracy on documents with high confidence.
Privacy and security
Invoice photos are stored only on your server. The Claude Vision API receives only the image bytes for data extraction. Original photos are deleted after 30 days. No document is sent to any third party outside the extraction process.