Request for proposal (RFP) documents describes a client's business and service requirements in natural language. RFP packages typically consists of tens to hundreds of documents each ranging from tens to hundreds of pages and come at variety of formats and structures. Processing RFPs manually is a tedious, error prone and slow process. It is a competitive advantage to a service provider to be able to process RFP documents to automatically extract client requirements, and understand how these requirements map to the internal offerings, products or solutions of the business to improve the efficiency of preparing RFP responses, and conduct sizing and pricing of the solutions. However, this is a challenging task due to the complexity and variety of forms that requirements are expressed, their level of detail, style and language expression. In this paper, we present a novel cognitive solution that employs linguistic-based and machine learning methods for automated processing of RFP documents for extracting requirement statements, and mapping them to offering taxonomies. We also present RFPCog as an interactive and explorative tool for analysis, refinement and browsing of requirements-offering mapping. The presented methods have been applied on RFPs submitted to a large IT service provider company, and the result of evaluation of the methods and tool by practitioners shows the effectiveness of the tool for intelligent requirement extraction and analysis.