AI terms explained in plain English
Thirty-six definitions for people who make decisions about AI: what each term means, one concrete example, why it matters for your business and how we use it in practice.
- 36 terms
- Reviewed by practitioners
- Standards named, dates checked
Short, conservative, useful
Every entry follows the same shape, so you can read one in two minutes or skim ten in ten.
Each term gets a one-sentence definition, two or three paragraphs in plain English with a concrete example, a short list of why it matters commercially, and a note on how Teksmobile uses it in our AI services. The twelve most important mechanisms also carry a small diagram. Related terms link to each other, so you can follow a thread from RAG to chunking to reranking without leaving the glossary.
Definitions are deliberately standard. Where a term refers to a specific standard, law or product (the OWASP Top 10 for LLM Applications, the EU AI Act, ISO/IEC 42001, Model Context Protocol, Salesforce Agentforce, the llms.txt proposal) we name it and avoid figures we cannot verify. Engineers from our AI practice review entries as models and regulations change; each page shows its last update. For depth, the Insights guides go further on most of these topics.
All terms, A to Z
A
Agentforce (Salesforce)
Salesforce's platform for building and running AI agents that work on CRM data and take actions inside Sales, Service and other Salesforce clouds.
ReadAI Agent
A system in which an LLM plans and executes multi-step tasks by calling tools, observing results and deciding what to do next, with defined limits.
ReadAI Governance
The policies, roles, controls and records an organisation uses to make sure AI systems are built and operated responsibly, lawfully and with accountability.
ReadC
Chunking
Splitting long documents into smaller passages before embedding them, so each piece can be retrieved and fitted into a model's context on its own.
ReadContext Window
The maximum number of tokens a model can take in at once, including instructions, conversation history, retrieved documents and the answer it is generating.
ReadE
Embeddings
Numeric vectors that represent the meaning of text, images or other content, so that similar items end up close together and can be compared mathematically.
ReadEU AI Act
The European Union's regulation on artificial intelligence, which classifies AI systems by risk and sets obligations for providers and deployers according to that risk.
ReadEvals (LLM Evaluations)
Repeatable tests that measure how well an LLM application performs on a defined set of inputs, so changes to prompts, models or data can be judged objectively.
ReadF
Fine-tuning
Further training of a pre-trained model on your own examples so it adopts a specific style, format or task behaviour without starting from scratch.
ReadFunction Calling
An API feature in which an LLM returns a structured call (function name plus JSON arguments) that matches a schema you declared, instead of free text.
ReadG
Generative Engine Optimization (GEO)
Making a brand's content more likely to be cited and summarised accurately by AI answer engines such as ChatGPT, Perplexity and Google AI Overviews.
ReadGuardrails
Controls placed around an LLM, on its inputs, outputs and actions, that enforce policy, block unsafe content and keep the system within its intended scope.
ReadH
Hallucination
A confident, fluent LLM output that is factually wrong or unsupported by any source, produced because the model predicts plausible text rather than checking truth.
ReadHybrid Search
A retrieval approach that combines keyword (lexical) search with vector (semantic) search and merges the two result lists into one ranking.
ReadI
Inference
Running a trained model to produce an output from new input, as opposed to training; in LLM systems it is where latency, throughput and most running costs appear.
ReadISO/IEC 42001
The international standard for an AI management system (AIMS), which specifies how an organisation should govern, risk-assess and continually improve its use of AI.
ReadL
LLM (Large Language Model)
A neural network trained on very large amounts of text that predicts the next token, which lets it generate, summarise, classify and reason over language.
Readllms.txt
A proposed convention for a Markdown file at the root of a website that gives AI systems a concise, curated guide to the site's most important content.
ReadM
MCP (Model Context Protocol)
An open standard, introduced by Anthropic, that defines how AI applications connect to external tools, data sources and prompts through a common client–server interface.
ReadModel Routing
Automatically sending each request to the most suitable model, cheap or capable, local or hosted, based on the task, its difficulty, cost limits or data rules.
ReadMultimodal AI
Models and systems that take in or produce more than one type of data, such as text, images, audio and video, within a single interface.
ReadP
Prompt Engineering
Designing the instructions, examples and context given to an LLM so it produces reliable, well-formatted results for a specific task.
ReadPrompt Injection
An attack in which untrusted text, typed by a user or hidden in a document or web page, makes an LLM ignore its instructions and do something its operator did not intend.
ReadR
RAG (Retrieval-Augmented Generation)
Retrieving relevant passages from your own documents and giving them to an LLM, so answers are grounded in your data rather than the model's memory.
ReadRed Teaming (AI)
Structured adversarial testing in which specialists deliberately try to make an AI system fail, leak data or misbehave, so weaknesses are found before attackers do.
ReadReranking
A second-pass step in retrieval that re-scores a shortlist of candidate documents with a more accurate (and more expensive) model, so the best matches rise to the top.
ReadRLHF
Reinforcement learning from human feedback: a training stage in which human preference ratings steer a language model towards helpful, honest and harmless responses.
ReadS
Semantic Search
Search that matches on meaning rather than exact keywords, usually by comparing the embedding of a query with the embeddings of documents.
ReadSpeech-to-Text (STT)
Technology that converts spoken audio into written text, also called automatic speech recognition (ASR), used for transcription, captions and voice interfaces.
ReadStructured Data
Machine-readable markup, usually Schema.org vocabulary in JSON-LD, that tells search engines and AI systems exactly what a page is about.
ReadT
Temperature
A sampling setting that controls how random a model's output is: low values make answers more predictable, high values make them more varied.
ReadTokens
The units of text, roughly word fragments, that a language model reads and writes, and the unit in which context limits and API prices are measured.
ReadTool Use
The ability of an LLM to request that the surrounding application run a function, API call or query, then use the returned result to continue its answer.
ReadFrom definitions to decisions
When a term turns into a project, these are the next steps.
Insights and guides
Long-form, opinionated guides on RAG, security, costs, the EU AI Act and more.
Browse guidesAI services
Strategy, LLM applications, agents, MCP servers, evals and monitoring, delivered by senior engineers.
See servicesAI readiness assessment
A ten-minute interactive check of where your organisation stands before you commit budget.
Take the assessmentFrequently asked questions
Who is this glossary for?
Business and technical leaders who need to follow an AI conversation, read a proposal or question a vendor without a research degree. Each entry is written for a reader who is clever but new to the term.
How are the definitions kept accurate?
Every entry uses conservative, standard definitions and names the relevant standard or product where one exists. Engineers from our AI practice review entries when models, standards or regulations change, and each page shows its last update.
Can I suggest a term?
Yes. Send us the term and where you met it. If it comes up in real projects, we add it.