Core Insights - Brex, a financial technology company, is aggressively implementing AI for internal operations, achieving nearly complete process automation [1][4] - Founded in 2017, Brex has raised over $1.5 billion with a valuation of $12.3 billion, and aims to exceed $500 million in revenue by 2025, paving the way for an IPO [2][3] AI Strategy - Brex's AI strategy consists of four parts, focusing on maximizing internal AI capabilities and treating the internal AI platform as a product [4][5] - The company aims to enhance operational efficiency by 5 to 10 times through AI, fundamentally changing its growth scaling law [4][37] Internal AI Platform - An internal platform was built to allow employees to safely and freely create, test, and deploy AI agents, addressing slow response times [9][10] - The platform is maintained by a systems engineering team and integrates AI capabilities from external products, creating a closed-loop product system [13][14] Workflow Redesign - Brex has restructured its operational workflow by shifting from task management to managing AI agents, allowing AI to handle repetitive tasks [15][16] - Employees are now expected to manage AI performance rather than just human tasks, requiring new skills in prompt engineering and data analysis [18][19] Training and Recruitment - A comprehensive AI training program was implemented for all employees, with assessments based on AI proficiency levels [20][21] - Recruitment strategies shifted from seeking specialists to looking for generalists with AI capabilities, including practical assessments of AI usage [24][25] AI Workflow Implementation - The company emphasizes understanding AI's advantages and redesigning workflows accordingly, focusing on tasks that are time-consuming and where AI can outperform humans [26][27] - Brex aims for quick wins in AI implementation, avoiding the pursuit of 100% automation to ensure practical and immediate benefits [28] Successful AI Applications - AI has shown unexpected effectiveness in subjective tasks, such as KYC processes, achieving an accuracy rate of 88% compared to human analysts [31] - In complex rule applications, AI has significantly reduced processing time from three hours to three seconds [34] - For clear step processes, AI has automated parts of the KYC workflow, freeing human resources for higher-value tasks [36]