An environmental consulting services company has a large staff of expert consultants and project documents spanning many decades. Each project document is a PDF of potentially several hundred pages. As they make proposals in response to various RFPs, they need the ability to search the repository of project documents for relevant prior work, extract and summarize relevant information, and synthesize it into a new document. Currently, this process depends on memory among long-time employees to find relevant project work, and manual work to read, extract, and summarize the information. Each new RFP may require a slightly different document processing flow over the same data.