We're paying OpenAI, Cohere and Textract separately just to parse and search our own docs

I work with one of Andrew NG's companies around document space. Our document pipeline hits three vendors for one flow: Textract for OCR, OpenAI for embeddings, Cohere for reranking. Three bills, three SDKs, three sets of rate limits, and the data leaves our cloud three times. Trying to work out whet