Glossary

11 terms

AI Agent

An AI-enabled software capability able to reason about objectives, interact with tools or systems, and perform actions within defined boundaries.

AI Architecture

The structures, components, relationships, principles and decisions required to integrate AI capabilities into the enterprise.

AI Governance

Decision rights, policies, controls, accountability and oversight mechanisms governing AI.

AI Readiness

The extent to which the enterprise possesses the capabilities necessary to adopt and scale AI for a defined purpose.

Enterprise AI

AI capabilities integrated with enterprise information, systems, processes, governance and operating models to produce organizational outcomes.

Fine-Tuning

Adaptation of an existing model through additional training for particular behaviors, domains or tasks.

Human-in-the-Loop (HITL)

An architecture pattern in which human participation is required for review, approval, correction or decision-making at defined points.

Model Context Protocol (MCP)

An integration protocol that may connect AI applications with tools, resources and context; AI4EA treats it as an implementation option.

Retrieval-Augmented Generation (RAG)

An architecture pattern in which external information is retrieved and supplied as context to a generative model.

Responsible AI

Principles, practices and controls intended to ensure AI is developed and used consistently with organizational, ethical, legal and societal expectations.

Use Case Priority Score (UCPS)

The AI4EA decision-support mechanism for evaluating opportunities using value, alignment, feasibility, readiness, risk and time to value.