Description
Summary We're a global brand looking for a developer to build an internal AI agent that turns our pile of contracts into one searchable knowledge base our team can actually ask questions of. Right now the process is all manual and has too many steps. Someone opens each contract, reads it, pulls out the key terms by hand, and files it away. When a question comes up later, we're back to opening files one by one to find the answer. It's slow, it's expensive at our scale, and it means nobody really has a clear picture across everything we've signed. We want an AI agent that reads every contract for us and builds a knowledge base we can search two ways. The first is by meaning, so someone can ask "which of our vendor agreements auto-renew in the next 90 days" and get the right contracts back. The second is the part that matters most to us, and the part most tools skip: we want to count and add things up across the whole collection. So a question like "how many of our contracts mention alcoholism" or "how many times does an indemnity clause show up" should give us a real number, not just a list of documents to go read. We want to ask a question of the whole library at once and get an actual answer. What the agent needs to do Read every contract and pull out the key details: parties, dates, payment terms, renewal clauses, obligations, and specific terms or topics we care about Let us search by meaning and get the right contracts back Let us count and total things across all contracts, so "how many times did X come up" returns a number we can trust Write a clear plain-English summary of any long agreement Keep everything in one place the whole team can reach Handle PDFs, scanned documents, and Word files without us reformatting anything first Who we're looking for Someone who has built AI document tools before and can show us real examples, especially anything where people could ask questions across a large set of documents and get answers back. We're a non-technical team, so we want a partner who explains the choices in plain words and hands us something our people can use on day one without a manual. It's a plus if you've worked with larger organizations and understand that we want something practical and dependable that fits how our team already works. Tech we're open to We trust your judgment on the stack. We know tools like Anthropic's Claude API and OpenAI are common for this kind of work, along with OCR tools like AWS Textract or Google Document AI for reading scanned files, and vector databases for the meaning-based search. For the counting side we'd expect some structured storage behind it, but we'll leave the exact approach to you. Tell us what you'd recommend and why it fits a job this size.