Plain-language explainers

Every rule, standard and technical term used on this site, explained in ordinary words. The pages on rules and standards link the official text at the end, and those links open in a new tab so you keep your place here. The pages on ideas and terms explain a concept rather than a document, so there is nothing to link out to.

New to this? Start with these five

They are the terms that come up first, and each one takes a couple of minutes. The full library is underneath whenever you want it.

Private deployment2 min read
Model gateways3 min read
Grounded answers2 min read
Baseline2 min read
The EU AI Act3 min read

Rules and standards

3 min readThe EU AI ActWhat Europe's AI law asks for, and who it reaches
2 min readISO/IEC 42001The management standard for how an organisation governs AI
2 min readThe NIST AI Risk Management FrameworkA voluntary US framework: govern, map, measure, manage
3 min readData protection law and AIWhy where the model runs, and what it reads, decides your obligations
2 min readModel risk managementThe discipline regulated firms use to keep models honest, applied to AI
3 min readAI rules around the worldA short map of the main laws and frameworks outside the EU

Ideas and terms

2 min readBaselineMeasuring how the work runs before anything changes
2 min readControl groupThe fair comparison that turns an impression into a number
2 min readThe AI profit and loss viewThree columns of value against the full cost
2 min readTotal cost of ownershipEverything an AI system costs, not just the licence
2 min readAI inventoryKnowing what you run before anyone asks
2 min readLanguage models and open weightsWhat a language model is, and what it means to own one as a file
2 min readPrivate deploymentRunning the model where you control it
3 min readModel gatewaysOne door every AI request goes through
2 min readOwning or renting the hardwareHow to decide, and why the answer is measured rather than assumed
2 min readTokens and inferenceThe unit AI is measured in, and why sizing depends on it
2 min readGrounded answersAnswers built from your documents, with the source attached
2 min readAI agentsSoftware that carries out a task and stops for a person
2 min readHuman sign-offWhere a person decides, and why it is designed in from the start
2 min readAccess and identities for AIGiving an AI system its own login, permissions and record
2 min readTuning and routingSmall models for routine work, large ones for the hard cases
2 min readEvaluation setsA fixed set of real questions that scores every change
2 min readSystem of recordThe one system that owns a fact

Next steps

Read the research
See how we help
Ask us a question