Overview · Glossary
AI and software terms, in plain language.
Short, practical definitions — what each term means, when it matters to you, and the mistakes to avoid — with links to the guides that go deeper.
AI
GlossaryRAG (retrieval-augmented generation)RAG is a technique where an AI system first searches your documents or data for relevant passages and then lets a language model answer using them, with cita…GlossaryEmbeddingsEmbeddings are lists of numbers that represent the meaning of text, so a computer can find passages that mean similar things even when they use different words.GlossaryVector databaseA vector database stores embeddings and finds the most similar ones quickly, which is how RAG systems locate relevant passages among millions.GlossaryFine-tuningFine-tuning is further training of an existing model on your own input–output examples to make its behaviour — format, style, classification — more consistent.GlossaryPrompt engineeringPrompt engineering is designing the instructions, examples and context given to a language model so it performs a task reliably.GlossaryAI agentAn AI agent is a system where a language model plans and takes actions through tools — searching, updating records, sending messages — to complete a task.GlossaryTool callingTool calling (function calling) lets a language model request that your code run a defined function — such as “look up order” — with structured arguments.GlossaryMCP (Model Context Protocol)MCP is an open protocol that standardises how AI applications connect to tools and data sources, so a system exposed once can be used by many AI apps.GlossaryHallucinationA hallucination is when a language model states something false or unsupported with confidence — an invented fact, citation, policy or number.GlossaryGuardrailsGuardrails are the checks and limits around an AI system — on inputs, outputs and actions — that keep it within its intended scope and policies.GlossaryLLM evalsLLM evals are automated tests for AI behaviour: a set of realistic inputs with expected results, scored on every change to prompts, models or data.GlossaryContext windowThe context window is the maximum amount of text (measured in tokens) a model can consider at once — instructions, history, retrieved documents and its own a…GlossaryTokensTokens are the chunks of text a language model reads and writes; they are how context limits are measured and how most model APIs are billed.GlossaryPrompt cachingPrompt caching lets a model provider reuse the processing of an identical prompt prefix across requests, reducing cost and latency for repeated content.GlossaryHuman in the loopHuman in the loop means a person reviews, approves or corrects AI output at defined points before it takes effect.
Delivery & contracts
GlossaryProof of concept (PoC)A proof of concept is a small, time-boxed build that tests whether a risky idea works — on real data, against criteria agreed in advance.GlossaryMVP (minimum viable product)An MVP is the smallest version of a product that real users can use to solve a real problem, built to learn whether the product should exist.GlossaryDiscovery phaseA discovery phase is a short, fixed-scope stage before development that defines goals, users, scope, architecture, risks and a reliable estimate.GlossaryChange requestA change request is a documented proposal to change agreed scope, with its cost and schedule impact estimated before anyone approves it.GlossaryTime & materials (T&M)Time and materials is a contract model where you pay for the time actually spent by the team, typically at agreed rates, with flexible scope.GlossaryFixed priceFixed price is a contract model where the vendor delivers an agreed scope for an agreed amount, carrying the estimate risk.GlossaryVendor lock-inVendor lock-in is when switching away from a supplier would be so costly or risky that you effectively can’t — because they control your code, infrastructure…GlossaryTechnical debtTechnical debt is the future cost of shortcuts taken in software today — code, architecture or process choices that make later changes slower and riskier.GlossaryCI/CDCI/CD (continuous integration and continuous delivery) is the automated pipeline that builds, tests and deploys software every time code changes.GlossaryInfrastructure as code (IaC)Infrastructure as code means defining servers, networks, databases and permissions in version-controlled files (for example Terraform or Bicep) instead of cl…
Want a second opinion on your project?
Tell us what you’re building and where you’re stuck. We’ll reply within one business day with the most practical next step — even if that step isn’t us.