Project: Digital Hive

The client managed influencer marketing campaigns across Xiaohongshu (Little Red Book), Douyin, and WeChat ecosystems. A 15-person operations team manually scouted KOLs, negotiated contracts, collected contact details, and generated campaign reports—burning 2,000+ hours per month on repetitive browser-based workflows.

We architected "Digital Hive"—a full-stack AI employee platform powered by a skill-driven ReAct agent engine. Each digital employee operates with its own persona (SOUL/IDENTITY), long-term memory, and a Tauri-based local browser agent that autonomously navigates live platforms like 蒲公英 (Dandelion). Natural language commands trigger a SkillRouter that dispatches specialised SKILL.md definitions to an AgentEngine, which drives real Chrome sessions via WebSocket. We deployed 11 distinct employee archetypes—from KOL scouts and contract negotiators to ad-plan auditors and livestream scriptwriters—all orchestrated through a single chat interface.

The 15-person ops team was replaced by 3 human supervisors overseeing the digital workforce. KOL screening throughput increased 800%. Campaign launch cycles compressed from 2 weeks to 48 hours. Total operational cost reduced by 70%, with zero data leaving the client's infrastructure.