Invoice processing in Odoo with AI: OCR, rules and learning from corrections

Odoo can read supplier invoices, but OCR alone is rarely enough: names are written differently, accounts and VAT codes repeat, and the accountant still fixes the same thing. The solution is an extra AI layer that learns from your practice.

How it works, step by step

• The OCR result is stored as a raw record, so everything can be traced and reprocessed.
• Rules engine: your explicit rules (for example, a given supplier always gets the same expense account and VAT) are applied before AI.
• Memory and aliases: the system remembers how a supplier or product was matched before and applies it to the new invoice.
• An AI model is called only when rules and memory are not enough: it proposes the account, tax and category and tidies the lines.
• Confidence score: every decision has a score. High-score invoices can be posted automatically, lower ones go to review with a visible explanation.

Learning from corrections

When an accountant fixes an auto-filled bill, the difference is saved and next time the same supplier is handled correctly. Within a few weeks manual corrections start to drop.

Why traceability matters

Each invoice has a log: which rules applied, what AI proposed and why, and the confidence score. That makes the system auditable and trustworthy.

Special cases

Extra rules can be added too, for example matching fuel or mobility card numbers to vehicles and their project analytics.

Who it suits

• companies receiving many supplier invoices a month;
• when invoices arrive in different formats;
• when accountants repeat the same corrections;
• when you want to see what was posted and why.

We built this AI layer ourselves and adapt it to your rules. For a general description of AI invoice recognition see “AI invoice recognition”. To see how it would work with your invoices, book a free consultation.