AI Business Professional key terms

Computer systems that perform tasks normally requiring human judgment, such as recognizing patterns, understanding language, or making recommendations, by learning from data or following encoded rules rather than only executing fixed instructions.

A branch of AI in which software improves at a task by finding patterns in example data instead of being explicitly programmed with rules for every case.

AI systems that create new content such as text, images, audio, or code in response to a user prompt, typically built on large models trained on broad collections of data.

A generative model trained on very large text collections to predict and produce language, enabling it to draft, summarize, translate, and answer questions in conversational form.

The trained mathematical structure that maps inputs to outputs, produced by exposing an algorithm to data; it is the artifact that makes predictions or generates content when run.

The examples used to teach a model its patterns. Its quality, coverage, and representativeness strongly shape what the model can do well and where it fails.

The act of running a trained model on new input to produce an output, such as a prediction or generated answer, as opposed to the earlier process of training the model.

A small unit of text, roughly a word fragment, that a language model processes and bills against. Usage and cost are usually measured in tokens rather than words or characters.

The maximum amount of text, measured in tokens, that a model can consider at one time, covering the prompt and its response; content beyond this limit is not available to the model.

A confident but incorrect or fabricated output from a generative model, such as an invented citation or statistic, arising because the system predicts plausible text rather than verifying facts.

AI built to perform a specific, bounded task such as spam filtering or demand forecasting, in contrast to hypothetical general intelligence; nearly all business AI today is narrow.

The instruction or input a user gives a generative AI tool, including any context, examples, or constraints, which strongly influences the quality and relevance of the output.

The practice of designing and refining prompts, using clear instructions, examples, and structured formats, to get more accurate, consistent, and useful responses from a generative model.

A prompting technique that includes a small number of worked examples in the prompt so the model can infer the desired format, tone, or reasoning pattern before handling the real request.

A technique that asks a model to work through intermediate reasoning steps before giving a final answer, which often improves results on multi-step problems but does not guarantee correctness.

Background instructions set by the application or administrator that define a model's role, tone, and boundaries before user input arrives, shaping behavior across a whole conversation or tool.

An approach where a system first retrieves relevant documents from an approved knowledge source and supplies them to the model, so answers are grounded in specific, current, verifiable content.

Tying a model's output to trustworthy, identifiable sources such as internal documents or databases so users can check claims and the system is less likely to invent facts.

Further training of an existing model on a smaller, targeted dataset to adapt its behavior or style to a specific domain, task, or organizational vocabulary.

The degree to which data is accurate, complete, consistent, timely, and fit for its intended use; poor quality reliably degrades AI outputs and the decisions built on them.

Information organized into a defined format such as tables, rows, and columns, making it easy to query, aggregate, and feed into analytics and many AI models.

Information without a predefined tabular schema, such as emails, documents, images, audio, and chat logs; it holds much business knowledge but requires processing before analysis.

The policies, roles, and controls that determine who may access, change, and share data, and how its quality, lineage, and compliance are maintained across an organization.

A record of where data came from, how it has been transformed, and where it flows, which supports troubleshooting, audit, and trust in AI inputs and outputs.

The automated sequence of steps that moves and transforms data from sources to storage and consumption, including extraction, cleaning, and loading for analytics or model use.

The principle that flawed, biased, or incomplete input data produces unreliable outputs, so data preparation deserves serious effort before any model is deployed.

When a model learns its training examples too closely, including their noise, and performs poorly on new, unseen data because it has not captured the underlying general pattern.

A numeric representation of text, images, or other content that captures meaning, allowing a system to compare items by similarity for search, clustering, or recommendation.

A data store built to hold embeddings and quickly find items with similar meaning, commonly used to retrieve relevant documents for search and retrieval-augmented generation.

Systematic unfair skew in data or model outputs that disadvantages particular groups or reflects historical inequities, producing inaccurate or discriminatory business results.

The degree to which people can understand why an AI system produced a given output, which matters for trust, debugging, regulatory compliance, and accountability in decisions.

Openness about how an AI system works, what data it uses, and its limitations, so users and stakeholders can make informed judgments about relying on its outputs.

A design in which a person reviews, approves, or corrects AI outputs before they take effect, used where errors carry significant legal, financial, or safety consequences.

Data that can identify a specific individual, directly or in combination with other data, such as names, contact details, or identification numbers, subject to privacy obligations.

The privacy principle of collecting and sharing only the personal data actually needed for a stated purpose, reducing exposure and limiting what AI systems can inadvertently reveal.

Technical and procedural controls that constrain what an AI system may accept or output, such as content filters, allowed-topic limits, and escalation rules, to reduce harmful or noncompliant behavior.

The structures, policies, and oversight an organization uses to ensure AI systems are lawful, ethical, secure, and aligned with business objectives throughout their lifecycle.

An approach to building and using AI that balances business value with fairness, safety, privacy, transparency, and accountability to affected people and stakeholders.

The potential for loss or harm from an AI model being wrong, misused, outdated, or poorly understood, which organizations manage through validation, monitoring, and controls.

The gradual decline in a model's accuracy as real-world conditions or data patterns change after deployment, requiring monitoring and periodic retraining or replacement.

A plan that connects AI initiatives to organizational goals, defining priority use cases, required capabilities, investment, governance, and how success will be measured.

The practice of ranking potential AI projects by expected business value, feasibility, data readiness, and risk, so limited resources go to the most worthwhile opportunities first.

A measure comparing the financial benefit of an initiative with its total cost, used to justify AI spending and to compare competing projects on a common basis.

The full lifetime cost of an AI solution, including licensing, integration, data preparation, training, support, and ongoing monitoring, not just the initial purchase price.

Using software, including AI, to execute repetitive steps in a business process with less manual effort, ideally freeing people for judgment-intensive and relationship-focused work.

Analyzing system event logs to reconstruct how a business process actually runs, revealing bottlenecks, rework, and automation opportunities that documentation may not show.

An AI assistant embedded in a productivity or business application that helps users draft, summarize, search, or analyze within their existing workflow rather than in a separate tool.

AI systems that can plan and carry out multi-step tasks, calling tools or other systems with limited supervision, which raises new questions of control and accountability.

The structured effort to prepare people for new processes and tools, addressing skills, incentives, and resistance so that AI adoption actually sticks and delivers value.

The baseline understanding of what AI can and cannot do, its risks, and how to use it appropriately, needed by employees to work effectively and safely with these tools.

The goal that AI systems treat individuals and groups equitably, without unjustified disparate outcomes, assessed through defined metrics and stakeholder input.