A junior data scientist is building a system to analyze legal documents. Their goal is to automatically determine which parties are bound by the contract, the key dates mentioned, and the legal jurisdiction. They propose building a single, large model that reads the entire document and directly outputs these three pieces of information in a summary. As a senior team member, critique this 'end-to-end' approach. Argue for or against an alternative strategy that would first involve a dedicated step to identify and label all person names, organization names, dates, and locations throughout the text before attempting to determine their specific roles in the contract. Justify your position by explaining the potential advantages or disadvantages of separating these tasks.