The AI Impact Communications Model has three parts: nine constructs you specify, six non-negotiables every service must meet, and three evidence layers holding it all together. Week 1 introduces them properly. This is the orientation view.
Select a construct to see what it covers and when you specify it in the course.
How your service shows its working to the audiences it serves. Source attribution, AI-assistance disclosure, abstain-and-escalate logic, complaint workflows. The operating layer that shapes every other construct.
Specified in Week 6Consent-first, layered, reversible tailoring, with a high-quality non-personalised path always available. Tailoring is opt-in, purpose-specific, and logged.
Specified in Week 8Curated narrative assets and the rules that govern them. Facts-and-options panels, risk-and-action pairing, AI-generated visual and audio governance.
Paired with multi-platform in Week 7Equivalence of meaning across channels. Same meaning, different form. Headless content patterns, accessibility parity at display, AI-mediated search adaptation.
Paired with storytelling in Week 7Equity-aware segmentation, frequency caps, and discretion-based delivery. Where commercial digital logic collides most directly with high-stakes communication ethics.
Paired with strategic repetition in Week 9Spaced reinforcement keyed to journey stage rather than engagement, with stop rules and fatigue monitoring built in.
Paired with targeted distribution in Week 9The governance backbone. Planning spine, canonical topic register, prompt governance, brand voice specification, journey map.
Specified in Week 13Connection without exposure. Pseudonymous participation, consent-first visibility, cultural safety verification, moderation including disclosure risk.
Specified in Week 10Indicator architecture, dashboards that drive action, drift detection, feedback loops, crisis communication protocol. How the service stays good after launch.
Specified in Week 11Audiences can see who wrote it, when it was reviewed, and what version they are reading.
Tailoring is opt-in, purpose-specific, reversible, and logged. A strong non-personalised path always exists.
The content meets accessibility standards in the channel where the audience actually sees it.
When the system is unsure, out of scope, or seeing risk signals, it pauses and hands off to a human.
Any risk message is paired with something the audience can actually do, and a route to support.
Every material change has a record, an audit trail, and a way to roll back.
Three bodies of evidence hold the framework together. You will not specify these directly. Week 3 walks you through them.
What we know about how people actually change behaviour. The behavioural-evidence anchor that runs through every construct's design rules.
Covers frameworks such as COM-B, the Behaviour Change Wheel, and the Theoretical Domains Framework.
How messages get processed, accepted, or rejected. The communication-evidence anchor.
Covers the Elaboration Likelihood Model, narrative transportation, fear-appeals and risk-action pairing, and acceptance models such as TAM and UTAUT.
The regulatory and standards environment your service has to live within. The governance-evidence anchor.
Covers NIST AI RMF, ISO/IEC 42001 and 42005, the EU AI Act, sector-specific frameworks, and country-specific guidance such as Australia's October 2025 Guidance for AI Adoption.