Automation glossary

AI Reasoning

AI reasoning is a model’s ability to work through a problem in steps before answering; reasoning models spend extra computation thinking through a task before they respond.

In plain language

Instead of answering immediately, the model plans, checks its work, and then gives the final answer.

Example

A reasoning model compares three vendor contracts clause by clause before recommending which one fits a policy.

Large Language Model (LLM) · AI Agent · Evals · Token (AI)

Common misconception

More reasoning is not always better; it adds cost and time, and simple extraction or classification rarely needs it.

Implementation guidance

Use reasoning for multi-step analysis and planning, keep fast models for simple steps, set budgets for thinking time, and evaluate whether it improves results.

Primary references

← Browse every glossary term