List of AI Terms - Demystifying AI - The Messy Middle Matters

	A	B	C	D	E	G	H	I	J	K
1						Mapping Terms to Proficiency
2	Term	Basic Description	More Technical Description	Relevant for Answering	Related to	Still Figuring it Out	Capable & Learning	Confident & Capable	Fluent & Building w/ AI	Pioneers of AI

3	AI agent	An AI system that can take actions and carry out multiple steps toward a goal.	A system that uses a model to select actions, invoke tools, and use the results to determine subsequent steps toward a goal, within configured permissions and controls.	How does AI become part of doing work?	AI workflow, Tools / tool use, Permissions, Instructions
4	AI provider	A company or organization that supplies AI models or services.	An organization that develops, hosts, or distributes AI models or services. The provider of an application may differ from the provider of its underlying model.	What am I actually using?	Model, App / interface
5	AI workflow	A sequence of tasks or steps that uses AI to help get work done.	A process that combines model outputs with human actions, software, or tools. Its steps may be predefined or dynamically selected, and it does not necessarily require an AI agent.	How does AI become part of doing work?	AI agent, Tools / tool use, Skill, Verification
6	App / interface	The software or interface through which you interact with AI.	The application layer through which a user accesses AI capabilities. It can add instructions, tools, information sources, and controls around an underlying model.	What am I actually using?	AI provider, Model, Instructions, Tools / tool use
7	Context	The information available to an AI as it works on a particular response or task.	The material supplied to a model for a particular generation, including instructions, conversation history, documents, and tool results. It is distinct from information learned during training.	What am I giving it to work with?	Prompt, Instructions, Context window, Tools / tool use
8	Context window	The limit on how much information a model can work with at one time.	The maximum number of tokens a model supports within a single interaction, subject to its input and output limits. Long conversations or documents may need to be shortened, summarized, or selectively included.	What am I giving it to work with?	Context, Tokens
9	Evaluations / Evals	Systematic checks of how well an AI system performs a task.	Structured assessments using defined tasks, criteria, and scoring methods to measure qualities such as accuracy, reliability, or task completion, and to compare systems or changes over time.	How should I judge what comes back?	Verification, Hallucination, AI workflow
10	Generative AI / GenAI	AI that can create content such as text, images, audio, video, or code.	AI systems that generate outputs based on patterns learned during training, guided by the instructions and information provided when they are used.	What am I actually using?	Model, LLM
11	Hallucination	When AI produces information that is incorrect or unsupported but presents it as valid.	Generated content that is fabricated, factually inaccurate, or inconsistent with the evidence available for a task. Fluent wording and confident presentation do not establish its accuracy.	How should I judge what comes back?	Verification, Evaluations, Context
12	Instructions	Directions that tell an AI what to do or how to respond.	Directions supplied by a user or built into an application that guide a model’s behavior, constraints, and output. Systems may assign different priorities to different sources of instructions.	What am I giving it to work with?	Prompt, Context, Skill
13	LLM / Large Language Model	A type of AI model that works with language, including generating and processing text.	A language model trained on large amounts of data to process and generate sequences of tokens. It can support tasks such as writing, summarization, translation, and coding.	What am I actually using?	Model, Generative AI / GenAI, Tokens
14	Model	The underlying AI system that processes information and produces a result.	A system whose parameters are learned through training and used to generate outputs or make predictions from inputs. A model is one component of an AI product.	What am I actually using?	Generative AI / GenAI, LLM, AI provider, App / interface
15	Permissions	The actions and information an AI system is allowed to access or use.	Access controls that determine which resources a system can read or modify and which actions it can execute. These controls are enforced by the surrounding software and services.	How does AI become part of doing work?	AI agent, Tools / tool use, App / interface
16	Prompt	The request or input you give an AI to guide its response.	Input supplied to a model, which may include a question, task, instructions, examples, or supporting information. It is processed alongside other available context.	What am I giving it to work with?	Instructions, Context
17	Skill	A reusable set of instructions that guides an AI through a particular kind of task.	A packaged procedure that an AI system can load and follow, sometimes including reference materials, examples, scripts, or tools. The exact format and behavior vary across systems.	How does AI become part of doing work?	Instructions, Context, Tools / tool use, AI workflow
18	Tokens	Small units of information that an AI model processes, such as words or parts of words.	Units used to represent model inputs and outputs. In text, a token may be a word, part of a word, or punctuation. Token counts help determine context limits, usage, and often cost.	What am I giving it to work with?	LLM, Context window
19	Tools / tool use	Capabilities an AI can use to do things such as search, calculate, or work with files.	Functions or external services that an AI system can invoke to retrieve information or perform actions. Tool results may be returned to the model as context for further work.	How does AI become part of doing work?	AI agent, AI workflow, Permissions, Context
20	Verification	Checking whether an AI’s answer or completed work is accurate and meets the request.	Checking specific claims, outputs, or reported actions against reliable sources, explicit requirements, tests, or observable results rather than relying on the system’s own assurance.	How should I judge what comes back?	Hallucination, Evaluations, AI workflow
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100