Resources · Glossary
Enterprise AI glossary
Plain-English definitions of the terms that come up in every AI programme. Share it with anyone joining the conversation.
- Agentic AI
- AI systems that pursue a goal by planning steps and taking actions in other systems, such as updating a record or drafting a document, rather than only answering questions.
- AI agent
- A piece of software that uses a language model to decide and carry out actions within set permissions. An agent has a job, tools it can use, and limits.
- AI agent vs automation
- Automation follows fixed rules written in advance. An agent handles variation, such as reading an unfamiliar contract, and decides the next step within limits. Most enterprise workflows use both.
- Human in the loop
- A control where a person reviews or approves an agent action before it takes effect, usually for actions above a set risk or value.
- Risk tier
- A level given to each agent based on what it can do and what could go wrong, which decides its approvals, monitoring and evidence.
- Audit trail
- A record of every agent action: what acted, on which record, with what input, under which policy version, and who approved it.
- Kill switch
- A tested way to pause an agent quickly, with a named person allowed to use it and a plan for work in progress.
- AI governance
- The policies, roles and controls that decide which AI is used, by whom, for what, and how its results are checked.
- Model risk
- The risk of loss or harm from a model that is wrong, misused or drifts over time. Banks have managed it for decades; agents extend it to more teams.
- Retrieval-augmented generation (RAG)
- A technique where the model looks up your own documents or records before answering, so responses are based on your data rather than only the model's training.
- Grounding
- Tying an agent's output to a source it can cite, such as a policy clause or a CRM record, so a reviewer can check it.
- AI inventory
- A register of every model and agent in use, with owner, purpose, data used, risk tier and status. Several regulators now expect one.
- AI literacy
- The knowledge staff need to use AI safely in their role. Required for organisations in scope of the EU AI Act since February 2025.
- High-risk AI
- Under the EU AI Act, uses such as hiring, credit scoring, education and critical infrastructure that carry extra obligations for documentation, oversight and accuracy.
AI transformation for the agentic engine era
Start with a fixed-fee AI transformation review. We map your systems, data and teams against the six layers of Pegasus Agentic OS and name the first workflows to hand to agents. You keep the findings whether or not you work with us.