Defined term
Chain of thought
Prompting the model to show intermediate reasoning steps before producing a final answer.
Chain-of-thought prompting asks the model to write out its reasoning step by step before giving a final answer. It improves accuracy on math, logic, and multi-step problems. Modern frontier models do this automatically via extended thinking modes. We surface chain-of-thought traces in audit logs for high-stakes decisions.
When it matters
When the task involves multi-step reasoning, math, or logical inference. Adds latency and cost but improves accuracy on hard tasks. Skip on simple classification.
Real example
A contract-extraction task where the model must compute prorated amounts. With CoT prompting ('Show your work step by step'), accuracy on the math jumped from 72% to 94%. Cost: +30% tokens; tradeoff: easy win at this volume.
KPIs to watch
Accuracy lift on hard subset (target: >10pp), token overhead per call (typically +20-50%), routing decision (use CoT only when complexity warrants).
Related terms
ReAct
A reasoning pattern where the model alternates Thought → Action → Observation steps.
Extended thinking
A model mode that performs longer internal reasoning before producing the answer.
Agentic AI
AI systems that can plan, take multi-step actions, and use tools to complete tasks autonomously.
Autonomous agent
An AI agent that completes a defined task without per-step human input.
See it in action
We use this every week
Book a 30-min call and we'll walk you through how Chain of thought shows up in a real engagement we're running.
Book a 30-min call