AI agent
An AI agent is software that can take actions inside your tools, not just answer questions. It can capture details, follow rules, hand off edge cases to a human, and complete repeatable work like follow-up, scheduling, and reporting.
Glossary
Clear definitions, written for operators, not engineers.
LLM (Large Language Model)
An LLM is the “language engine” behind many AI systems. It reads and writes text (and sometimes images) and is used for tasks like summarising, extracting details, classifying messages, and drafting responses.
Workflow automation
Workflow automation is rules-based automation that moves data between tools (for example, “when a form is submitted, create a CRM record and send an email”). It’s reliable for structured triggers, but it doesn’t handle messy edge cases well.
RAG (Retrieval-Augmented Generation)
RAG is a pattern where an AI system searches your approved knowledge (docs, FAQs, policies) and uses those results to answer more accurately. It reduces hallucinations by grounding responses in your real information.
Human-in-the-loop
Human-in-the-loop means the agent can do the routine work, but a person approves sensitive actions (like refunds, price changes, or sending legal language). It’s the safest way to automate high-stakes steps.
Guardrails
Guardrails are constraints that keep an agent reliable: what it’s allowed to do, what it must never do, when it should ask clarifying questions, and when it must escalate. Guardrails are a core part of production deployments.