Framer AI AgentsDesign and publish professional sites with AI
Tax Radar
AI-powered NCM tax classification for Brazil's tax reform
5 followers
AI-powered NCM tax classification for Brazil's tax reform
5 followers
Tax Radar uses AI to classify products under Brazil's NCM tax code from a free-text description — no barcode (EAN) required, unlike most tools on the market. It automatically flags differentiated tax treatment under Brazil's new Tax Reform (IBS/CBS, LC 214/2025) — zero-rate and 60%-reduction categories — with a confidence score and explanation. It also audits NF-e and SPED files to catch classification errors before they become compliance risks.
Brazil's new Tax Reform (IBS/CBS) is forcing companies to review how their products are classified for tax purposes. Getting the NCM, CST or cClassTrib wrong can mean applying the wrong tax treatment across an entire product catalog.
Most tax classification tools depend on a barcode/EAN lookup. But many Brazilian catalogs, especially B2B and industrial products, do not have reliable EAN data.
That is why we built Tax Radar: an AI-powered NCM classification tool trained on Brazilian fiscal data. It works directly from free-text product descriptions, returns the most likely NCM with a confidence score, and helps identify whether the product may qualify for IBS/CBS tax treatments such as zero-rate or reduced-rate regimes.
Happy to answer questions about the ML approach, NCM classification, cClassTrib, or Brazil's tax reform itself.
How does it actually handle the free-text input when someone describes a product with super generic terms like "industrial part" or just drops a brand name with no real description?
@asoglubirs45419 Great question. We added an input-quality checker that runs while the user types. It looks at things like character count and whether the description includes both a noun and an adjective, so if someone enters something too generic like “industrial part” or only drops a brand name, the UI gives visual feedback asking them to describe the product in more detail.
If the input is still vague, Tax Radar doesn’t pretend to be certain. It returns a lower-confidence result and explains what extra context would improve the classification, such as material, function, sector, or intended use. The goal is to make uncertainty visible instead of overclaiming.