AI Basics & Models
This is the broad field of making computers perform tasks that typically require human intelligence, like perception, reasoning, or decision‑making.
What is Artificial Intelligence (AI)
This is a central repository that stores raw, large‑volume data from many sources in its original format—structured and unstructured—for analytics and AI.
What is a data lake
This is the act of crafting the instructions or input we give a generative model to get a desired response.
What is a prompt
This happens when an AI model confidently returns an incorrect or fabricated answer, especially in generative systems.
What is an AI hallucination
In business AI, this describes a defined business problem and value statement, such as “reduce call center handle time by 10% using an AI assistant."
What is a use case
This term describes an AI tool embedded in an application—like email or CRM—to help users draft content, summarize, or automate tasks.
What is a copilot or AI assistant
This subset of AI uses data and algorithms so that systems can learn patterns and improve performance on tasks without being explicitly programmed.
What is Machine Learning
This newer architecture combines the flexibility of a data lake with the management and performance features of a data warehouse.
What is a data lakehouse
This discipline focuses on designing, iterating, and optimizing prompts (and sometimes tools/context) to improve AI output quality.
What is prompt engineering
This refers to methods like better retrieval, tighter prompts, and validation checks that reduce incorrect or made‑up outputs from generative models.
What is hallucination mitigation
This small‑scale experiment is designed to prove that a concept is technically feasible, often before investing heavily.
What is a POC (Proof of Concept)
These are numeric representations of text, images, or other objects in a high‑dimensional space, enabling similarity search and powering many RAG systems.
What are embeddings
This type of AI focuses on creating new content—like text, code, images, or audio—rather than just classifying or predicting outcomes.
What is Generative AI
This term describes the end‑to‑end system that turns raw data into AI‑powered outcomes, usually including data pipelines, feature stores, models, and deployment.
What is an AI factory
This approach augments generative models by first searching external data sources and then feeding the retrieved context into the model along with the user’s request.
What is RAG (Retrieval‑Augmented Generation)
These are rules, constraints, and safety controls—such as content filters, policy checks, or tool limits—that define what an AI system is allowed to do or say.
What are guardrails
This limited rollout puts a solution into a real environment with real users to validate performance, adoption, and business value at small scale.
What is a pilot
These policies, roles, and controls define how data is managed, secured, and used appropriately across its lifecycle.
What is data governance
This type of model is trained mostly on text and can understand and generate human‑like language at scale, including powering many copilots and chatbots.
What is a Large Language Model (LLM)
This kind of database stores numerical representations of text, images, or other objects to support similarity search for things like RAG or recommendation systems.
What is a vector database or vector store
These operational practices apply DevOps‑like principles to the lifecycle of machine learning models and AI systems, including monitoring, deployment, and updates.
What are MLOps or AI Ops
These practices and policies ensure AI is fair, safe, transparent, and aligned with human values, and that risks like bias and misuse are proactively managed.
What is Responsible AI or Trustworthy AI
Over time, this phenomenon occurs when a model’s performance degrades because the real‑world data it sees changes compared to the data it was trained on.
What is model drift
This broader discipline ensures that AI initiatives follow organizational policies, regulations, and ethical guidelines, often using committees and formal processes.
What is AI governance
In an AI workflow, this phase is where the model is used to make predictions or generate outputs on new data, as opposed to learning from training data.
What is inference
This is the process of moving and transforming data from sources into usable formats for analytics and AI—often including ingestion, cleaning, and feature creation.
What is a data pipeline
These very current AI systems can break tasks into multiple steps and call tools or APIs—such as databases or business systems—instead of only answering questions.
What is agentic AI or agentic RAG
These legal and policy concepts govern where data is stored and processed, often requiring that certain data remain in specific countries or regions.
What are data residency and data sovereignty
These related concepts describe how well humans can understand why a model made a particular prediction or generated a particular output.
What are explainability and interpretability
This training‑time vs. deployment‑time distinction refers to updating model parameters using data vs. using a fixed model to answer queries in production.
What are training vs. inference