OCR messy invoices without per-vendor templates

Rule-based (IF/THEN) tools come from the 1990s: you set up each vendor by hand before they will read anything, and you tell them which fields to grab. untxt. does the opposite.

The 1990s way (rule-based templates)

• Configure every vendor before it reads a single invoice.
• Map fields, pick dropdowns, tag the vendor, set the category.
• A new vendor or a redesigned invoice breaks the rule.
• It captures only the fields you pre-mapped, and you maintain all of it forever.

untxt.

• Starts reading the moment you upload it.
• Reads the whole document and presents what is actually on it: vendor, dates, totals, tax, and every line item, with tables kept intact.
• Shows you everything it finds, so you review what is there instead of choosing fields up front.
• Handles a brand-new vendor on the first invoice exactly like the thousandth.

As bookkeepers put it, they spend more time writing rules than reviewing data. With untxt., the work is reviewing what it found.

This is exactly where rule-based tools fall over and untxt. holds up: dozens of clients, hundreds of suppliers, every one with a different layout. There is nothing to configure per supplier. A new format is simply another document. That is what makes it workable for a practice adding clients every month instead of babysitting a fixed vendor list.

Understanding is not magic. A genuinely chaotic layout, a faded thermal receipt, or handwriting will come through with lower confidence, and untxt. flags those values for you to confirm rather than posting a confident guess. You stay in control of what is uncertain, and it will not invent a number to fill an empty field.

Per-vendor templates were always a workaround for tools that could not read context. They are a maintenance tax dressed up as a feature. The first problem is understanding the document, and building rules is not something you should have to do to get there.