How Edda knows
Most school AI tools are a chat window over a pile of documents: they retrieve something that looks relevant and write a confident answer. That is fine until the answer goes in front of a leadership team, a parent, or an inspector. Edda is built differently, because her answers have to survive exactly those rooms.
The knowledge base
Edda's knowledge is not scraped or summarised. Every standard is held word for word, and every relationship between them is an explicit, recorded link.
The regulator's AI-literacy framework, the OECD and European Commission's AILit framework (the basis for PISA 2029), ISTE's student and educator standards, UNESCO's three frameworks for AI and ICT competency, AI4K12, and the IEEE digital-intelligence standard. Every indicator from every one of them is either mapped into Edda's combined framework or built into it, and the coverage is verified item by item, not claimed in a total.
A combined progression of over 440 student indicators from early years to Grade 12 across five domains, each carrying what emerging, meeting, and exceeding look like, assessment guidance, rubric criteria, and adaptations for different learners. Alongside it, a teacher framework of 119 indicators by capability level. All of it aligned to the IB programmes a school actually teaches.
When a framework changes, Edda's base changes with it: when the AILit framework moved from draft to final, every affected indicator, mapping, and document was re-derived against the published version, and the differences were traced item by item. Answers never lean on a superseded draft.
Citations, not confidence
When a teacher asks what the regulator requires for Grade 6 AI ethics, Edda does not paraphrase from memory. She resolves the question to the real indicators, quotes them exactly, and names the source document. If something is not covered by any framework, she says so, rather than inventing a plausible answer.
Each code and clause in an answer points at a real entry in the knowledge base. Before any release ships, an automated test bank of realistic teacher and coordinator questions runs against the system, and a single unresolvable or unsupported citation blocks the release.
The test bank includes the questions designed to catch a system lying: standards that do not exist, competences removed from a framework's final version, plausible codes one digit off. The required answer is an honest no, with the real alternatives shown.
The same questions run through an ordinary retrieval system score markedly lower on accuracy and honesty. We keep the scorecard, and we would rather show it than describe it.
Your school, on the same map
The knowledge base is the shared layer. When a school connects its curriculum, its own units, policies, and delivery records join the same map, so a question like "which of my Year 5 units covers AI ethics, and did we deliver it?" gets an answer with the same discipline: named units, real dates, traceable links. That chain, from requirement to indicator to unit to what actually happened, is what an inspection asks for, and it assembles itself.
See it yourself
The fastest way to test a system that claims it can show its working is to ask it something you already know the answer to. Bring your hardest framework question to a demo.