AI playbook glossary

16 terms · Ilura Technology

WHAT THIS IS Short, honest definitions for the terms used across ILURA Academy — from playbooks and learned preferences to App Intents and the Data Not Collected label. Each term has a stable anchor you can link to.

Personal agent

A user-trained assistant behavior for a role, routine, or recurring task. In ILURA, agents are invoked by the user, reviewed by the user, and learn from saved corrections.

Decision rule

A saved criterion that tells an agent how to choose between options, such as when to be firm, when to offer an alternative, or what next step must stay visible.

Routine

A repeated workflow an agent can help structure when invoked, such as a weekly update, follow-up, support reply, or note-to-message pattern.

Playbook

A saved, reusable set of writing instructions — tone, rules, and examples — that an AI applies consistently and that you invoke by name.

Learned preference

A correction you saved. When you fix an output and keep the fix, the playbook records it as a rule it will follow from then on.

Style memory

The accumulated set of learned preferences inside a playbook. The more you correct, the more the playbook sounds like you on your best day.

Training by correction

Teaching an AI by editing its output and saving the edit, instead of writing longer instructions up front. Each saved correction becomes a learned preference.

On-device AI

AI that runs entirely on your phone or computer. Your text is processed locally and never travels to a server.

Apple Intelligence

Apple's AI system built into iPhone, iPad and Mac. Its core models run on the device itself, which is what makes private text generation possible.

App Intents

The Apple framework apps use to expose actions to Siri, Shortcuts and the system. It is how an app becomes addressable by voice.

Share Sheet

The iOS share panel that appears when you tap Share. Apps can register actions there, so you can send selected text straight into a tool without copy-paste.

Siri Shortcut

A saved action or chain of actions you can run by voice or one tap. Shortcuts can pass text into apps and bring results back.

Role-based AI

Using a separate, specialised assistant voice per job — Sales, Manager, Founder — instead of one generic assistant that writes everything the same way.

Re-prompting

Repeating the same instructions to an AI in every new chat because nothing you taught it earlier persists. The tax that saved playbooks remove.

Data Not Collected

The strictest App Store privacy label. It means the developer collects no data from the app at all — no analytics, no identifiers, no accounts.

Version history

A numbered record of how a playbook changed as it learned. You can review what changed between versions and roll back at any time.