Abstract: With the increasing complexity of supply chains in the power and engineering industries, traditional authorized procurement processes have exhibited inefficiencies and inaccuracies in bill preparation, price limitation, and risk management. To address these issues, this paper proposes an intelligent risk control method for the entire authorized procurement chain based on large-scale models. The method constructs a comprehensive risk assessment model, a dynamic price-limiting control function, and a self-learning correction mechanism to enable multidimensional data fusion and intelligent analysis across the procurement stages of "service bill-price limit-authorization-performance". A pilot study conducted on a power supply company's supply chain test platform demonstrates that the proposed approach significantly improves risk identification accuracy and price rationality, shortens approval cycles, and effectively reduces performance deviation rates.