Sensible reads the paperwork nobody wants to - loss runs, rent rolls, bank statements, a fax of a fax - and hands your software clean, auditable JSON.
Somewhere right now, an account manager at an insurance brokerage is squinting at a scanned ACORD form, retyping numbers into a web app one field at a time. This is the un-glamorous engine room of the modern economy - and it is exactly where Sensible has quietly set up shop.
Mature industries love to talk about digital transformation. Insurance, real estate, lending, logistics, healthcare - trillions of dollars move through these sectors every year. And yet the actual interface, the thing where the data lives, is still a document: a PDF, an email attachment, a photograph of a signed page. The software can't read it. So a human does. Sensible's founding observation is almost rude in its simplicity: the document is an API that nobody bothered to write.
Co-founder Josh Lewis learned this the hard way. Before Sensible, he was building internal tools at Newfront Insurance, wrestling quotes, applications, and certificates of insurance into something a computer could use. He went looking for a developer tool that sat one honest layer above the raw OCR dump - flexible, controllable, trustworthy - and found nothing. So in 2020, with co-founder Ming Lu, a product leader out of Lattice and Intercom, he built it.
There are two schools of document extraction. Rigid rule-based systems are exact but shatter the moment a layout changes. Large language models are flexible but will happily, confidently, invent a number that was never on the page. Sensible's answer is not to pick a side - it runs both, then wraps the whole thing in guardrails.
Upload a file, hit the API, or simply forward an email attachment.
LLM techniques plus layout-based rules pull out rows, tables, checkboxes, and sections.
Schema enforcement, confidence scores, and source coordinates for every field.
Clean JSON lands in your product - with a full audit trail and optional human review.
Developer-first endpoint that turns PDFs, scans, images, and emailed attachments into structured JSON.
A query language blending LLMs with visual layout rules to extract rows, tables, checkboxes, and sections.
Bank statements, pay stubs, ACORD forms, driver's licenses, closing disclosures - solved out of the box.
Not just typed text - charts, checkboxes, and handwriting become data too.
Confidence scores, source coordinates, audit trails, and human-in-the-loop review.
Forward a document to Sensible and get structured data back. The inbox becomes a data pipeline.
Teams across insurance, proptech, fintech, logistics, and healthcare wire Sensible into their products so their people stop being human data-entry machines. A sampling:
Josh Lewis and Ming Lu found Sensible after Lewis's document-parsing frustration at Newfront Insurance.
Pre-seed backing from Engineering Capital; early developer platform and prebuilt parsers take shape.
$6.5M seed round led by Craft Ventures - "make documents as accessible to software as APIs."
Hybrid LLM + layout-based engine and an expanded parser library push deeper into insurance, proptech, fintech, logistics, and healthcare.
Return to that account manager and the scanned ACORD form.
In the version of the story where Sensible is wired in, the squinting and the retyping simply don't happen. The form arrives, the API reads it, the fields land in the system - each one carrying a confidence score and a pointer back to the exact spot on the page it came from. If something looks off, a human glances at a flagged field instead of transcribing all forty of them. The paperwork still exists. The drudgery around it does not.
That's the quiet ambition here. Sensible isn't trying to replace anyone's software - it's trying to feed it, turning the documents at the edge of every business into data the software can finally use. The most boring object in the building, made legible to machines. Not flashy. Just, well, sensible.