The language of agentic work.

Clear definitions, practical examples and the distinctions that matter.

20 terms to explore

A–Z

Agent-native: Describes a system or service designed with software agents as primary users or participants.
Agent-ready: Describes a system that supports a specified interaction by a software agent.
Example & source
A supplier allows an agent to check stock, while ordering still requires a person.
Always state ready for what. Supporting one lookup does not establish end-to-end agent access or an agreed readiness level.
What is Agent-Native?
Builder.io, Agent-Native
One implementation of agent-native application design; not a universal definition.
Agentic: Describes AI software that chooses and carries out steps towards a goal rather than following only a predetermined sequence.
Agentic commerce: Buying and selling in which software agents participate in commercial choices or actions for people or organisations.
Agentic economy: Economic activity in which AI agents help people and organisations produce goods, deliver services, coordinate work, and buy and sell.
AI: The broad field of systems that infer from information how to produce predictions, content, recommendations or decisions towards objectives.
Example & source
A factory might use AI to flag an unusual equipment reading; a writer might use it to draft an explanation.
AI includes more than generative AI or agents. Using AI does not by itself mean a system chooses and carries out a sequence of work.
Explanatory memorandum on the definition of an AI system
OECD · 2024
Plain-language interpretation of the institutional definition.
AI agent: An AI system that can select steps towards a goal and carry out actions through available tools.
AI assistant: An AI application through which a person asks for help with information or work.
Example & source
A coordinator asks an assistant to draft a message explaining what a technician needs before a visit.
Assistant describes the relationship with the user. An assistant may contain agentic features; it is not a fixed level below an agent.
Levels of Autonomy for AI Agents
Kevin Feng, David McDonald and Amy Zhang · July 2025
A proposed research framework, not an adopted industry standard.
AI model: The learned component of an AI system that maps supplied information to predictions or other outputs.
Example & source
A model interprets a customer’s message; the surrounding application retrieves their booking and displays a reply.
The model is one component. Connections, permissions and the interface belong to the system around it.
Machine Learning Glossary
Google for Developers
Technical terminology, paraphrased for a business reader.
AI-native: Describes a product, service or business designed around AI as a central part of how it delivers value.
Augmentation: Use of AI to support a person’s work while the person remains actively involved.
Example & source
A trainer asks for alternative explanations, compares them and edits the one used in a workshop.
Augmentation and automation can coexist within the same job. Active involvement is not evidence that the AI is less capable.
Economic primitives
Anthropic · January 2026
Research distinguishes patterns of assistance, automation and autonomy; this is a general explanation, not an Index classification.
Authority: Permission to take specified actions on someone’s behalf, within stated limits.
Example & source
A coordinator permits an agent to request information, but reserves appointment changes for a person.
Technical access and available funds do not establish permission. Authority can be narrow even when a system is highly capable.
Explanatory memorandum on the definition of an AI system
OECD · 2024
Plain-language interpretation of the institutional definition.
Automation: Work carried out by a system with reduced human action, using fixed rules, AI or a combination of both.
Example & source
A booking system sends a reminder the day before each appointment without someone composing each message.
Automation does not require AI, and a highly automated task may leave the software very little discretion.
Economic primitives
Anthropic · January 2026
Research distinguishes patterns of assistance, automation and autonomy; this is a general explanation, not an Index classification.
Autonomy: The degree to which a system can learn or act without human involvement after people delegate work to it.
Delegation: Assignment of work to a person or system, together with the decisions they are permitted to make.
Example & source
“Prepare the visit from our existing records; ask me before contacting the customer” delegates work with a clear boundary.
Delegating work does not transfer all responsibility, grant unlimited access or remove the need to check the result.
Explanatory memorandum on the definition of an AI system
OECD · 2024
Plain-language interpretation of the institutional definition.
Generative AI: AI that creates synthetic content, such as text, images or audio, using patterns learned from data.
Example & source
A teacher asks for three practice questions based on an approved lesson and reviews them before use.
Generating material does not necessarily involve choosing a plan, using tools or acting on someone’s behalf.
Generative artificial intelligence
NIST · SP 800-218A
Paraphrase of the definition cited by NIST.
Service: Work performed for another party, such as transport, teaching, equipment maintenance or bookkeeping.
Example & source
A technician repairs equipment for a customer; an information provider supplies a requested record.
A service need not be digital or agentic; people can deliver it directly, with or without software.
Concepts and definitions for measuring business innovation
OECD / Eurostat · Oslo Manual 2018
Economic context for services; the definition here uses ordinary business language.
Task: A bounded piece of work contributing to a larger objective.
Example & source
Checking whether a booking has the equipment model number is one task within preparing a repair visit.
A task is not a whole job, occupation or industry. Completing it may help without achieving the larger objective.
Economic primitives
Anthropic · January 2026
Research distinguishes patterns of assistance, automation and autonomy; this is a general explanation, not an Index classification.
Tool: A function an AI system can use to obtain information or take an action.
Example & source
A calendar connection lets an agent check whether an appointment time is available.
A tool need not be paid or external. It supplies an action; the agent decides when to use it within its permissions.
Machine Learning Glossary
Google for Developers
Technical terminology, paraphrased for a business reader.
Workflow: A connected arrangement of tasks and decisions through which work proceeds.
Example & source
A repair workflow moves from receiving a booking through collecting details, scheduling a visit and recording what happened.
Workflow is broader than fixed automation. When comparing AI architectures, identify whether rules or a model choose the steps.
Building effective agents
Erik S. and Barry Zhang, Anthropic · December 2024
The workflow–agent distinction used here; terminology varies across the field.