What is it?
Effort scaling
Giving an agent explicit rules about how much work a task deserves, so it does not over-invest in easy questions or under-invest in hard ones. For example: a simple fact check should use one agent and a few tool calls, while a complex research task might justify ten or more sub-agents. Without such rules, agents tend to either spin in circles or quit too early.
From the AI agent glossary
Where Effort scaling is explained
- The Taxonomy of Agentic RAG SystemsAgentic RAG · Adaptive agentic RAG
- Practical Guidance: When and How to Use Agentic RAGAgentic RAG · Lesson 4: Agent autonomy requires explicit constraints
- The Case For Multi-Agent: Anthropic's Research SystemMulti-Agent Systems · Prompt engineering becomes the main control lever