Challenge
Challenge
Companies want useful AI in operations, but simply adding a chatbot to group chat creates scattered answers, weak access control, limited auditability, and high rollout risk.
AI Transformation / Lark
AI Control Center for a Lark-based company
Instead of demoing another chatbot, OPERCEL designed an AI control center model: employees work through Lark, agents run on OpenClaw, skills are managed in a registry, and memory and permissions are separated by role.
9capability groups
RBACaccess model
Skillscore primitive
Approach
Approach
- Use a skills-first model where each agent has a clear workflow, data boundary, and owner.
- Design Lark interface, agent runtime, memory graph, model routing, and audit trail together.
- Roll out by department instead of launching company-wide in one step.
Expected result
Expected result
- Leadership and departments get a system blueprint instead of a chatbot demo.
- Permission, memory, and model cost risks are handled early.
- The company has a practical pilot path before expansion.