Alkemia

Learn

What would you like to understand about AI?

Most AI explanations assume you already know the vocabulary. These do not. Pick the question closest to yours and start there — nothing here assumes a technical background.

What actually is AI, in business terms?

Start here if the vocabulary is the barrier. These are the ideas everything else depends on.

Definition

AI

Software that performs tasks normally requiring human judgement — in practice today, almost always a language model.

Definition

Large language model (LLM)

A model trained to predict the next piece of text, which turns out to be enough to summarise, draft, classify and reason about language.

Definition

Agent

An LLM given tools and a goal, which takes multiple steps on its own rather than answering a single question.

Definition

Chatbot

A conversational interface that answers questions but does not change anything in your systems.

Definition

Hallucination

When a model produces confident, fluent output that is simply wrong — the failure mode that makes AI risky in business processes.

Where should my business start?

The first decision is rarely which tool. It is which problem — and whether AI is even the right answer to it.

Definition

Proof of concept

A small build that answers a specific question about feasibility — not a small version of the finished product.

Definition

Pilot

A limited production deployment with real users, run to find out what actually happens.

What will it cost, and what do I get back?

AI has a variable cost per use, unlike most software. Understanding that early is what keeps a pilot from becoming a surprise.

Definition

Cost per call

What a single AI interaction costs you — the number that decides whether a feature is viable at scale.

Definition

Token

The unit models read and write, and the unit you are billed in — roughly ¾ of a word in English.

Definition

Model routing

Choosing which model handles a request at run time, usually by difficulty, cost or availability.

What can it actually do in my operations?

Concrete applications, and where each one tends to work or fall over.

Definition

RAG (retrieval-augmented generation)

Looking up relevant material from your own documents and giving it to the model, so answers are grounded in your data rather than its training.

How do I know it is getting things right?

The hardest question, and the one most vendors avoid. This is where the real engineering is.

Definition

Eval

A repeatable test of whether your AI system is actually getting things right — the difference between improving it and guessing.

Definition

Approval gate

A checkpoint where an AI-proposed action requires human sign-off before it executes.

Definition

Human in the loop

Designing the workflow so a person reviews or approves at the points where being wrong is expensive.

Definition

Failure taxonomy

A written, shared vocabulary for the specific ways your AI system goes wrong.

Definition

Guardrails

The constraints around a model that keep its behaviour inside acceptable limits.

Is my data safe?

What leaves your systems, who can see it, and what you should settle before a pilot rather than after.

Definition

Data governance

Knowing what data you hold, who may see it, and where it is allowed to go — the prerequisite most AI projects skip.

Definition

Training data

The text a model learned from — which determines what it knows, and when its knowledge stops.

Definition

Prompt injection

An attack where text the model reads contains instructions that hijack its behaviour.

Still stuck on a word?

The dictionary defines everything in plain English

Thirty-five terms that come up in AI decisions, each with a one-sentence definition and why it matters commercially.

Open the dictionary

Next step

Tell us what you’re trying to build

Most engagements start with a fixed-price audit, so the first thing you buy is a decision rather than a commitment.

Start a conversation