RunnelWorks

RunnelWorks

Bookkeepers spend a surprising amount of time on a deceptively simple task: figuring out what a document is and where it belongs. An invoice comes in as a PDF, a receipt as a phone photo, a statement as an email attachment. Each needs to be identified, categorized, and routed to the right place. Runnelworks is a document processing platform that automates this work. It uses OCR to extract text from uploaded documents, then applies vector embeddings and semantic search to classify and organize them.

The core insight is that financial documents cluster semantically even when they look different visually. An invoice from a restaurant supplier and an invoice from a tech vendor have different layouts, but their content shares structure—vendor name, line items, totals, dates. Embeddings capture this similarity in ways that rule-based systems can’t. The pipeline works like this: documents come in, OCR extracts the text, the text gets embedded, and a classifier determines the document type. From there, relevant fields can be extracted and the document routed to the appropriate workflow.

I’m building Runnelworks because I think there’s a lot of low-hanging automation fruit in back-office operations—work that’s repetitive and error-prone but hasn’t been worth the integration cost to fix. Modern ML tooling is finally making these problems tractable for small-scale, real-world use cases. Learn more at RunnelWorks.com.