Day 6: Planning in Agents + Reasoning Models

Today’s article focuses on the critical role of planning in building effective agentic AI systems. While agents can use tools and retrieve information, they often fail because they lack the ability to determine what to do next. Newer “Large Reasoning Models” (LRMs) are being developed to address this, trained specifically to plan and sequence actions – unlike traditional LLMs which primarily predict the next token – but they require careful implementation due to potential drawbacks like increased cost and overthinking.