Curriculum design is one of the most labour-intensive professional activities in education, and one of the least supported by software. The typical process for developing a new subject area curriculum in a large New Zealand secondary school involves a head of department consulting the New Zealand Curriculum framework, identifying relevant Achievement Standards from the NZQA database, mapping learning progressions across years, writing learning intentions for each unit, sourcing or creating assessment tasks, and formatting all of this into documents that can be shared with teachers, reviewed by senior leadership, and revised each year. Most of this work happens in word processing software. Some of it happens in cloud document tools. Almost none of it is stored in a structured format that could be queried, compared, or built upon systematically.
The curriculum design problem is not unique to New Zealand. Across the English-speaking world, curriculum development is a professional services market — staffed by experienced educators, curriculum developers, and instructional designers, producing deliverables that are primarily documents rather than software. Publishers and curriculum houses produce core curriculum resources, but the process of adapting those resources to a specific school's context, cohort, and strategic priorities remains manual and intensive. The total professional labour cost of this activity, across schools, districts, publisher teams, and ministry curriculum teams, is very large. And because it produces documents rather than data, almost none of that intellectual effort is accumulated in a form that can be built upon by the next person who does the same work in a different school a year later.
Generative AI changes this in two ways. The first is the obvious one: AI-assisted drafting substantially reduces the time required to produce a first draft of a learning sequence, a set of learning intentions, or a differentiated assessment task. A teacher who might spend two hours writing a unit overview from scratch can now produce a draft in fifteen minutes that incorporates relevant curriculum standard references, suggested learning progressions, and differentiated task options — and then spend the remaining time on the professional judgment work of refining that draft against her knowledge of her specific students. The time savings are real, and the productivity improvement for the curriculum writing function is significant. Pedagogue, which we backed in 2025, is building the tooling for this workflow: a generative AI curriculum platform designed specifically around the professional curriculum writer's process rather than the general-purpose document creation workflow.
The second change is more structurally significant and less visible in the near term: when curriculum design work happens in a structured digital environment rather than in Word documents, the outputs become data that can be accumulated, analysed, and built upon. A curriculum design platform that stores learning intentions as tagged objects, mapped to curriculum framework nodes, with version history and collaborative annotation, is not just a productivity tool — it's a curriculum knowledge base that compounds in value as more curriculum work happens inside it. The aggregate intelligence of thousands of teacher curriculum decisions, made explicit and machine-readable, is a training resource for future AI curriculum support that no individual institution could build on its own.
We should note the obvious tension here: curriculum is a domain where values, cultural context, and pedagogical philosophy matter deeply. The question of what should be in a curriculum — what knowledge is worth transmitting to young people, in what sequence, at what level of depth — is not a question with a technically correct answer. Generative AI that produces curriculum drafts will reflect the values and emphases of its training data, which means it will reflect the dominant curriculum traditions of its training sources. In a NZ context, that has specific implications: tools built primarily on US or UK curriculum content will not handle te reo Maori integration requirements or the specific cultural contexts of Pasifika learners in ways that NZ teachers would find satisfactory. The companies building AI curriculum tools for the Australasian market need to be building with NZ-specific curriculum frameworks and culturally contextualised content from the start, not retrofitting localisation later.