AI AGENTS FOR BUSINESS RESOURCE CENTER
Use this glossary to translate agentic-AI terminology into practical operational meaning.
Reviewed and updated July 2026.
Core concepts
- AI agent
- Software that pursues a goal by interpreting context and taking permitted actions.
- Agentic AI
- AI systems designed to plan, use tools, and act across multiple steps.
- Autonomy
- The degree to which a system can act without case-by-case human approval.
- Copilot
- An assistant that helps a person perform a task while the person remains in control.
- Orchestration
- Coordination of agents, workflows, tools, state, and handoffs.
- Multi-agent system
- A system in which specialized agents collaborate or divide responsibilities.
Models and knowledge
- Large language model (LLM)
- A model trained to process and generate language and other structured representations.
- Context window
- The information available to a model during one inference.
- Embedding
- A numeric representation used to compare semantic similarity.
- Retrieval-augmented generation (RAG)
- Retrieving relevant source material and supplying it to a model before generation.
- Grounding
- Tying output to authoritative data or verifiable evidence.
- Hallucination
- A plausible-sounding but unsupported or incorrect model output.
Tools and execution
- Tool calling
- A model selecting a defined function or API with structured arguments.
- API
- A defined interface through which software systems exchange requests and data.
- Workflow
- A sequence of tasks, decisions, and handoffs that produces an outcome.
- State
- Stored information about a workflow’s current progress.
- Idempotency
- Design that prevents repeated requests from causing duplicate effects.
- Human in the loop
- A design requiring human review, input, or approval at defined points.
Safety and governance
- Guardrail
- A control that limits output, data access, or actions.
- Prompt injection
- Malicious or untrusted content that attempts to redirect model behavior.
- Least privilege
- Granting only the minimum access required for a task.
- Audit trail
- A record of inputs, decisions, tool calls, approvals, and outcomes.
- Observability
- Logs, traces, metrics, and evaluations used to understand system behavior.
- Evaluation
- Systematic measurement of quality, safety, cost, and task performance.
- Escalation
- Routing a case to a person or controlled process when the agent should not proceed.
Business measures
- Containment rate
- The percentage of cases completed without human takeover.
- Correction rate
- The percentage of outputs a person must materially fix.
- Cost per outcome
- Total operating cost divided by successfully completed business outcomes.
- Payback period
- Time required for net benefits to recover the initial investment.
- Service-level objective (SLO)
- A measurable reliability or performance target.
Continue learning
- Complete guide to AI agents for business
- AI agent use cases by function and industry
- AI agent implementation framework
- Security and governance
- Cost and ROI planning
- Platform buyer’s guide
- AI agent glossary
Find the best AI agent opportunity in your business
Complete the guided business assessment to receive an AI Systems Business Analyst report, prioritized automation opportunities, and a recommended AI Systems Agent Team.