In this paper, we describe the natural language (NL) question- answering system for financial domains. Technique of semantic headers is ap- plied to represent the semi-structured and logically complex data in the form of textual answers by matching the semantic representation of a query with the ones of the answers. Multiagent architecture of financial advising is suggested, where each agent represents the specific domain coverage and viewpoint. We analyze the customer experience and knowledge engineering process for the Tax domain, which is rather sophisticated on one hand and requires rather precise answers on the other hand.