Unstructured @UnstructuredIO

"Mostly right" isn't good enough for an invoice. GPT-5.6 Sol read 96.1% of fields correctly on 200 scanned invoices. Only 1% of invoices came back with zero errors. Unstructured Transform: 84%. unstructured.io/blog/rag-neede…

Who you'll meet at Booth #1112 this week 👀 Come say hi! We're here all week 🦩 #GartnerSYM

Good enough parsing was fine when a human read the answer. Now an agent acts on it. @_Brian_Raymond on why the bar moved, and how to hold every parser to it 👇

Brian Raymond @_Brian_Raymond

6 days ago

The question I hear from engineering teams has changed, and I think it's the most important shift in the stack right now. In 2024 it was "why would I need the extra tenth of a percent?" Their parser did a pretty good job transforming a file into JSON or markdown, the retriever

0 0 0 272 2

We're all set up and ready to roll at @Gartner_inc IT Symposium! Come see us at Booth #1112. #GartnerSYM #EnterpriseAI #AgenticAI #GenAI #Unstructured

Gartner IT Symposium is all about separating AI hype from reality. Our corner of reality is the messy files your agents run on. Find us at Booth #1112 all next week. 🦩 #GartnerSYM

96.1% field accuracy sounds shippable. On 200 scanned invoices, that frontier model read 2 with zero errors. Transform: 99.4% on fields, 168 invoices clean. Tesseract: none. Agents don't consume averages. unstructured.io/blog/rag-neede…

RAG needed mediocre data. Agents need extraordinary data. Our new ebook on why one misread field matters a lot more once an agent acts on it. Includes a 200-invoice, 25-step test. 🔗 unstructured.io/blog/rag-neede…

Send us your ugliest pages. We mean it. 👀 Crooked scans. Coffee stains. Tables inside tables inside tables. Bundle up 25 and take the Hot Mess Test. We’ll put Unstructured Transform to work on them and show you how it stacks up against your current parser, side by side. Free. No setup. Results within 24 hours. Bring on the mess. 😎 🔗 unstructured.io/hot-mess-test

99% field accuracy sounds great. Until that 1% sends an entire form back for human review. That's why field-level accuracy doesn't tell the whole story. When evaluating document parsing, the question isn't just how many fields are correct. It's how many documents make it through without anyone needing to fix them. We put that to the test with 40 forms. Here's what we found: unstructured.io/blog/how-to-pa…

us jumping into the weekend like:

Every parser wins its own benchmark. We'd rather win yours. 💯 That's the Hot Mess Test. Send us your 25 hardest documents: the torn scans, the coffee stains, the forty nested tables, the handwriting in the margins. Unstructured Transform will likely get most of them right out of the box. Where it doesn't, our model team fine-tunes it directly on your files until it does 😎💅 - You send it: PDFs, scans, or images. Nothing to set up. - We run it, and tune where needed: your formats and edge cases, not a generic benchmark. - You compare: same file, side by side with your current parser's output. Free. No commitment. You keep the output either way. Take the Hot Mess Test: unstructured.io/hot-mess-test

Two parsers can post nearly identical field accuracy and still be miles apart in practice. Field accuracy counts fields. But a single wrong field is enough to send a whole form back for human review, so the number that matters is how many forms come back clean. In our test of 40 forms, Transform returned 34 with zero errors. The frontier model returned about half, at roughly 7.5x the cost per page. Cheaper to run, and more of your forms never touch a human. Read more in our latest blog: unstructured.io/blog/how-to-pa…

Brian Raymond @_Brian_Raymond

2 weeks ago

Field accuracy is the number everyone reports when they talk about parsing forms. It's also the one most likely to mislead you. We ran 40 grant applications through Unstructured Transform and GPT-5.6 Sol. On field accuracy they looked almost identical: 99.5% against 98.7%. You'd

0 0 1 282 0

A frontier model can match a parser field for field and STILL cost you more. We saw it in our own test. On individual fields, GPT-5.6 Sol and Unstructured Transform came out about even. But GPT got more whole forms wrong, and it was several times more expensive per page. Across thousands of documents, that means a bigger bill and more forms for your team to fix. We wrote about why, and what to check when you're putting a parser up against a frontier model: unstructured.io/blog/how-to-pa… #DocumentParsing #DocumentAI #DataEngineering #Unstructured

Brian Raymond @_Brian_Raymond

2 weeks ago

Last week we wrote about why better frontier models haven't made document processing go away. So this week we tested it, on forms. If you're evaluating a parser, field accuracy is the first number most people look at, and it's also the most misleading. We ran 40 grant

0 0 4 282 0

Forms look simple to a human. But they're not to an agent. A lot of what makes a form understandable actually has nothing to do with the words. Its the layout. Which label owns which value, which box is checked, which field is blank on purpose. Read a form like a paragraph and you keep the words and lose all of that. This week, were digging into why forms are deceptively hard to parse, what goes wrong when you treat them like plain text, and costs you when you treat them like flat text. Read the full breakdown: unstructured.io/blog/how-to-pa…

Wrapping up a very good week in Palo Alto 🫶 Chris had a great boardroom session at CxO Institute Palo Alto talking agent context, enterprise data, and what it takes to keep AI slop out of the picture. Thanks to everyone who hung out with us!

fall friday energyyyy 🍂 feat PSL obvi ☕

keyboard_arrow_left Previous keyboard_arrow_right Next

Read more on this in our latest blog 👀unstructured.io/blog/why-front…

Getting an answer from AI is one thing. Trusting it is another. 🫠 Reading a page has gotten easier. But getting every table, field, decimal, and source exactly right is still the hard part for models. And as agents take on more real work and this category keeps growing, that level of accuracy (and knowing you can trust it!) matters more than ever.

a productive afternoon for the queso committee 🙂‍↕️🌶️ (@Chilis we're your biggest fans)

keyboard_arrow_left Previous keyboard_arrow_right Next

We're all set up and ready for a great day at CxO Institute Palo Alto! We've got some great meetings lined up, and our Head of Product and Engineering @ctmaddock is breaking down how to unlock the knowledge trapped across enterprise data. Let’s get into it! 🏃‍♀️