Chain-of-Thought refers to a prompting technique in which a language model reveals its solution path in individual steps instead of just outputting the final result. A simple trigger is often enough: the instruction "Think step by step" is sufficient for the model to formulate intermediate steps before arriving at an answer.
The effect is noticeable in multi-step, logical or computational tasks. A model that is asked directly for the sum of ten items will make calculation errors more often than one that first lists each item individually and only adds them up at the end. The detour via the visible intermediate steps forces the model to check its own calculation before outputting it. For Users Does this have a secondary benefit: the chain of intermediate steps can be read and checked for plausibility before adopting the result. Anyone who wants to trace an AI-supported evaluation, such as a multi-stage plausibility check of a receipt, benefits from precisely this transparency.
Newer language models with built-in "reasoning" now carry out these steps automatically, even without an explicit instruction in the prompt. However, chain-of-thought remains the underlying concept: making intermediate steps visible instead of letting them disappear within the model.
Demarcation
Chain-of-Thought is by Few-Shot Prompting to distinguish them. Few-shot prompting shows the model examples of the desired outcome, while chain-of-thought shows the path to get there. Both techniques can be combined by ensuring the examples themselves already include worked-through intermediate steps. Tree-of-thought and self-consistency extend this basic idea by examining and comparing multiple chains of thought in parallel instead of relying on just a single one.
AI in everyday work: Here's how to find the right start
AI assistant meets SAP Business One: Two paths out of the Versino AI shop
Sprint and Swirl: How an AI agent divides work and learns from trials
LATS: Considering several paths simultaneously with AI — and the availability check in SAP B1
ReAct and RLEF: How Artificial Intelligence Learns from Real Mistakes — and What Automated Bank Reconciliation in SAP Business One Has to Do with It
