In line with analysts, Databricks and Snowflake’s choices would assist enterprises reduce down the complexity of workflows required to research unstructured information, particularly paperwork.
Enterprises, traditionally, have needed to construct complicated, gradual, brittle OCR pipelines in the event that they wish to deliver information from paperwork, equivalent to PDFs, into an AI workflow, ensuing within the end result of RAG, which enabled semantic search over parsed textual content however nonetheless struggled with nuanced doc constructions like tables, mentioned Bradley Shimmin, follow lead of information, analytics, and infrastructure at The Futurum Group.
To deal with paperwork with tables, enterprises typically chained extra LLM calls to extract and reconstruct tables as JSON, which was efficient however dangerous as a result of hallucinations, Shimmin mentioned, including that as an alternative of sewing collectively OCR, RAG, and customized extraction logic, Databricks’ ai_parse collapses your entire workflow right into a single declarative SQL assertion.
