Learning guide
AI Tooling Terms for Finance Readers
A beginner-friendly explanation of prompts, tools, agents, retrieval, evaluation, and guardrails.
AI terms are workflow terms
Many AI words describe how a system is built or checked. A prompt is an instruction. Retrieval brings outside information into the model context. Tool calling lets a model request a structured action. Evaluation measures whether the system behaves well enough for a task.
For finance readers, these words matter because AI tools increasingly appear in research, customer support, compliance, monitoring, and automation workflows.
Agents and tools
An agent is often described as a system that can plan, use tools, and continue across steps. Tool use can include search, calculation, database lookup, file reading, code execution, or workflow actions.
Useful agents need constraints. Guardrails, sandboxing, rate limits, and human-in-the-loop review can reduce the chance that automation performs an unsafe or unintended action.
Evaluation and monitoring
An eval suite tests model behavior against examples or scoring rules. Drift monitoring watches whether performance changes over time. Observability records traces, errors, tool calls, and user outcomes so a system can be improved.
These terms are not about hype. They are operational vocabulary for checking whether an AI system works reliably.
How this appears in the game
AI cards usually group by workflow: prompting, retrieval, tools, evaluation, monitoring, or safety. If a card describes how an AI system receives information, checks output, or calls an external tool, compare it with other automation terms.
The puzzle treats AI as a vocabulary domain next to crypto and finance because many modern market tools combine all three.