The slowest part of carbon accounting is not calculation — it's data entry. Hundreds of invoices, delivery notes and spreadsheets, each read and classified by hand. The Carbonlogy engine steps in right there.

The process has three layers. First, document scanning: uploaded invoices are converted into structured data through optical character recognition and layout analysis. Supplier, amount, unit, and date fields are extracted.

Next, automated classification: each line item is matched to an emissions category and an appropriate emission factor. The model draws on sector context and your previous corrections to sharpen its prediction.

The final layer is the most important: human oversight. The system assigns a confidence score to every match; low-scoring items are placed in a review queue. AI does not make the decision — it accelerates the decision and leaves an audit trail.

Why a human layer?
An audit-ready inventory must be able to show where every number comes from. Carbonlogy records every automated decision, so an auditor can trace it back step by step. Automation does not reduce transparency — it increases it.

— Burak, May 2026