Glossary
Plain definitions of the AI terms that appear in business cases, vendor proposals and board papers — with the misconceptions that cost money.
Updated 3 September 2026
An AI agent is a software system that pursues a stated goal over multiple steps, deciding which actions or tools to use along the way rather than following a fixed script. In business settings it is usually a language model connected to systems it can read from and write to, operating inside defined permissions with a human approval point for consequential actions.
Generative AI describes models that produce new content — text, structured data, code, images or audio — based on patterns learned from training data, rather than selecting from predefined options. In business use it most often means large language models applied to documents, correspondence and structured extraction.
AI readiness is the degree to which an organisation can deliver, operate and govern an AI-supported workflow. It spans six dimensions: data availability and quality, process stability, systems and integration access, skills, governance, and change capacity.
AI ROI is the net annual benefit of an AI-supported workflow divided by the total investment required to deliver and operate it: (annual gross value − annual running cost) ÷ first-year investment. Gross value is built from released labour time, avoided error and rework, and — where attributable — capacity or revenue effects.
AI automation is the use of models to perform workflow steps that previously required human interpretation — reading a document, classifying a request, drafting a response — usually combined with conventional automation for the deterministic parts of the same process.
An AI business case is the document that justifies investment in a specific AI-supported workflow. It states the measured baseline cost of the current process, the specific change proposed, the value range with confidence, the full first-year and run-rate cost model, the risk and governance position, the measurement plan, and a single decision request.
A large language model (LLM) is a model trained on very large text corpora to predict continuations of text. In practice this makes it able to classify, extract, summarise, translate and draft, given instructions and context, without task-specific programming.
Retrieval-augmented generation (RAG) is a pattern in which relevant passages are retrieved from a document or data store and supplied to a language model as context before it generates an answer. It grounds output in specific, current, owned content rather than training data alone.
AI governance is the set of policies, controls and accountabilities that determine how AI systems are approved, deployed, overseen, documented and monitored. In the EU it also covers classification and obligations under the EU AI Act, alongside GDPR duties for the data involved.
Merjora identifies where these ideas apply in your business, and what each opportunity is worth.
Discover your AI opportunitiesEditorial standard. Merjora publishes analysis, frameworks and publicly documented examples. We do not publish invented statistics, unattributed benchmarks or unverified customer stories.