AI GLOSSARY
AI glossary
Plain-language definitions of the AI terms business teams hear most.
A
- AI agent
An AI-powered assistant set up for a specific role. It uses the knowledge and tools you give it to answer questions and complete tasks.
Related: Knowledge base, Prompt
- AI governance
The policies, oversight, and controls a business uses to make sure AI is used safely, responsibly, and in line with its goals.
Related: Guardrails, Approved models
- AI model
The trained system that generates responses. Models from different providers vary in speed, cost, and strengths.
Related: Large language model (LLM), Approved models
- Approved models
The set of AI models a business allows its team to use, so everyone works with options the business trusts.
Related: AI governance, AI model
C
D
E
- Encryption
Scrambling data so only someone with the right key can read it. Encryption protects information in storage and in transit.
Related: Tenant isolation
G
- Grounding
Giving an AI model trusted source material, such as your documents, so its answers reflect facts from your business rather than general knowledge alone.
Related: Knowledge base, Hallucination
- Guardrails
Rules that shape what AI can do, such as blocking sensitive data or limiting which models can be used.
Related: AI governance, Data masking
H
- Hallucination
When an AI model produces an answer that sounds confident but is not accurate. Grounding answers in trusted knowledge helps reduce it.
Related: Grounding, Human in the loop
- Human in the loop
A practice where people review or approve AI output before it is used for important decisions.
Related: AI governance
I
- Integration
A connection between two software tools that lets information move between them, such as linking a workspace to Slack or Google Workspace.
Related: Workflow automation
K
- Knowledge base
The collection of documents, data, and resources an AI agent draws on to answer questions about your business.
L
O
- Operational intelligence
Using your business's own knowledge, activity, and data to make faster, better-informed decisions day to day.
Related: Knowledge base
P
- PCI data
Payment card information, such as credit card numbers, that must be handled under the Payment Card Industry Data Security Standard.
Related: PII, Data masking
- PII
Personally identifiable information: details like names, email addresses, or Social Security numbers that can identify a person.
Related: PCI data, Data masking
- Prompt
The instruction or question you give an AI model. Clear, specific prompts usually lead to better answers.
Related: AI agent, Context window
R
- Retrieval-augmented generation (RAG)
A technique where an AI system looks up relevant information from a knowledge source before answering, so responses are based on that material.
Related: Knowledge base, Grounding
- Role-based access control
Giving people access based on their role, so each person sees and does only what their job requires.
Related: AI governance
T
- Tenant isolation
Keeping each customer's data separate from every other customer's in a shared software platform.
Related: Encryption
- Token
A small unit of text, roughly part of a word, that AI models read and write. AI usage is often measured and billed in tokens.
Related: Context window
W
- Workflow automation
Using software to run a repeatable series of steps automatically, so routine work happens consistently without manual effort.
Related: Integration
GET STARTED
Ready to put AI to work?
Get started with Cohrt today, or subscribe to updates at the bottom of this page.