AI is easier to understand as overlapping capabilities, methods, system layers, and application domains than as one ladder from simple to intelligent.
Plan for this page
Understand the idea, test it actively, then explain it in your own words.
You will leave with
distinguish a capability, method, model, product, and application domain place one familiar system on more than one part of the map recognize when an AI/non-AI boundary depends on purpose, mechanism, or definition
Time
30 min
Before you begin
AI from Zero is useful preparation, but it is not a locked prerequisite. Each unit states the concepts it uses.
Do this now
Read the outcomes and useful terms before the first section.
distinguish a capability, method, model, product, and application domain
place one familiar system on more than one part of the map
recognize when an AI/non-AI boundary depends on purpose, mechanism, or definition
Terms you will use
These definitions prepare you for the reading; you do not need to memorize them.Capability
What a system does: perceive, classify, predict, generate, plan, optimize, recommend, or act.
Method
How designers build the capability, including explicit rules, search, statistical learning, neural networks, or hybrids.
System
The model or algorithm plus data, interface, tools, infrastructure, policies, people, monitoring, and consequences.
01
Maps answer different questions
A map of capabilities asks what a system can do. A map of methods asks how behavior is produced. A product map asks which components, people, data, and controls turn a model into a service. An application map asks where and why the system is used.
Confusion appears when one map is treated as all maps. ‘Generative AI’ describes a capability family, ‘machine learning’ a broad method family, ‘chatbot’ an interface pattern, and ‘healthcare’ an application domain. One product can occupy all four labels without making them synonyms.
02
The field is a web, not a clean tree
Vision may classify, generate, retrieve, or guide a robot. Language systems may translate, rank, extract, generate, call tools, or support decisions. Robotics combines perception, control, planning, hardware, and human environments.
Some capabilities use learned models; others use fixed rules or search. Many useful products are hybrid systems. Ask which component is learned, which is deterministic, which has access to external information, and which human institution defines the goal.
03
Boundary questions require a purpose
Whether something counts as AI can depend on a research tradition, law, procurement policy, product description, or ordinary conversation. A fixed tax formula may be automation without AI. A learned fraud classifier may be AI even if it produces no language.
Do not solve every disagreement with a label. State the definition or purpose, then describe the mechanism and consequence precisely enough for the decision at hand.
Interactive model
Four layers for locating a system
Selected layerInput and environment
Text, images, sensor signals, records, user behavior, physical space, or retrieved documents.
Read the complete text alternative
Input and environmentText, images, sensor signals, records, user behavior, physical space, or retrieved documents.
Capability and methodClassification, generation, planning, control, optimization, search, rules, learned models, or hybrids.
Product and workflowInterface, tools, data stores, approvals, monitoring, people, policies, and recovery.
Purpose and consequenceThe human goal, affected people, incentives, possible benefit, failure, and remedy.
Misconception to correct
“Every system that produces an output is generative AI.”
A calculator produces output, a classifier assigns categories, and a recommender ranks items. Generation is one capability family, not a synonym for all computation or AI.
Active checks
Decide, then compare the reasoning
No grade, score, or streak. Progress records completed learning objects, not your worth or ability.01
A phone app transcribes speech, searches a calendar, and suggests meeting times. Which description is most useful?
02
Two teams disagree about whether a fixed route-planning algorithm is AI. What is the strongest next move?
Explain back and transfer
Map one real system
Choose a familiar service. Describe its input/environment, capabilities and methods, product/workflow layer, and purpose/consequences. Mark anything you cannot determine from public evidence.
Answer both checks before completing.
Source context
What supports this unit and where it stops
Evidence status: established
Explanatory memorandum on the updated OECD definition of an AI systemOECD.AI
Useful for a system-level definition and policy context; definitions are purpose-bound and do not create one timeless boundary for every field.
inAi