Artificial Intelligence

Which AI concept must your team understand before it chooses a tool or control?

The Spine

• AI Glossary — establish shared terms for models, agents, memory, orchestration, and evaluation.
• How Models Learn — separate the narrow training objective, observed capabilities, and contested scaling hypothesis.
• How Models Think — distinguish fluent output from evidence of intact reasoning.
• AI Architecture — connect models, context, memory, agents, tools, and controls.
• Machine Learning — understand generalisation, neural networks, systems, and decisions.
• Context Graphs — map the decision context available to an agent.
• Models and Modalities — compare providers and the media they can interpret or generate.
• Prompting — direct intelligence with explicit intent and constraints.
• Agent Protocols — coordinate communication and transactions between agents.
• AI Agents — understand agent identities and workforce roles.

Zoom Out

Attention is the primitive. Human attention notices and values. Machine attention weights, relates, and predicts. Agent attention holds context long enough to act.

Apply these principles through AI-native work mapping, choose tools through the AI Toolkit, and inspect agent systems through AI Agents. Then follow the substrate into AI compute, AI data, and energy.

Sources

Visual maps	Current practice	Strategic frames
Diagrams · Matrices · Thinkers	Model guidance · Working practice	AI framing · Situational awareness

Context

• depends-on Decisions — names the judgment the selected AI concept must improve.
• applies-to Capabilities — turns principles into repeatable human and agent practice.
• pairs-with Protocols — connects probabilistic intelligence to coordination rules.
• risk-governed-by Software Algorithms — distinguishes probabilistic intelligence from deterministic infrastructure.
• applies-to Culture — keeps adoption grounded in the people and norms that carry it.
• pairs-with AI Overview — returns to the wider AI subject map.
• depends-on Software Platform — supplies the wider ABCD substrate for applied intelligence.
• applies-to Agentic Frameworks — turns architectural principles into agent-building choices.
• applies-to Culture — shows where adoption changes shared behaviour.
• pairs-with Protocols — routes into the wider coordination domain.
• applies-to Agency — asks which human intention the system amplifies.
• applies-to Problems — keeps AI selection anchored to a problem worth solving.

Questions

Which route would change your next AI decision?

Changes my mind: a cold reader cannot choose the right principle or application route without scanning the full section.

Next question: after learning the concept, which observable decision or control will you change?