Reported impact
Business outcome
ECLICKTECH moved beyond demo agents into a continuously running production system for advertising operations. The company says AI now covers about half of core execution work, single-operator efficiency has improved by more than 40%, diagnosis that once took hours can now be narrowed to minutes, and standardized skills plus a shared AgentCore base make it faster and cheaper to extend the model to additional business lines.
- Productivity Improvement
- Revenue Growth
- Risk Reduction
- Scalability
Solution
ECLICKTECH expanded its unified AI middle platform AdClaw with Alibaba Cloud AgentCore and AgentLoop to create an always-on ad-delivery agent factory. The architecture has three layers: business systems, a unified AI and data middle platform, and intelligent multi-cloud infrastructure. ACK runs the runtime environment, OSS stores shared data and assets, and ACR manages container images. On top of that base, skills, MCP connectors, knowledge services, and business tools become reusable capabilities, while an agent layer handles routing and task execution and a collaboration layer coordinates multi-agent flows plus human confirmation nodes. AgentCore provides model management, skill orchestration, MCP integration, AgentTeam collaboration, and credential governance. AgentLoop records model inference, tool use, RAG and memory activity, and sub-agent traces, then feeds Golden Set, Bad Case, Agent-as-a-Judge, and business-evaluator loops for continuous optimization.
Data flow
Business-system requests enter AdClaw, where AgentCore routes them into reusable skills, MCP connectors, knowledge services, and business tools running on the ACK-based platform. Shared data and assets are held in OSS and packaged workloads are managed through ACR. AgentLoop records model inference, tool calls, RAG or memory activity, and sub-agent traces, and the resulting evaluation data feeds Golden Set and Bad Case loops before optimized outputs return to advertising operators.
