Most EdTech products treat assessment as a feature: a quiz at the end of a module, a progress check built into the practice layer, a grade export that flows into a student management system report. This is understandable from a product development perspective — assessment is complex to build well, and most teams add it to satisfy a buyer requirement rather than because they have a considered architectural view of what assessment does in a learning system. The consequence is that assessment in most EdTech products is a reporting artefact rather than an instructional signal. It records what happened; it doesn't change what happens next.
The companies that treat assessment as a platform — as an API that other products build on top of — have a structurally different competitive position. The value of an assessment API is not in the items themselves; items can be licensed, written by curriculum teams, or generated at scale. The value is in the data model: the way items are tagged to curriculum standards, the calibration data that gives each item psychometric parameters, the response schema that allows downstream systems to act on student answers rather than just store them. When that data model is well-designed, every product that integrates the assessment API inherits a body of curriculum-aligned, psychometrically calibrated assessment infrastructure that would take years and millions of dollars to build independently. The API owner becomes infrastructure rather than application.
Learnosity, which we backed in 2024, is the clearest example of this model in the Australasian EdTech market. Their core product is an assessment question API — a set of item types, authoring tools, and response data schemas that EdTech platforms can integrate to deliver rich, interactive assessment experiences without building assessment from scratch. The customer list reads like a who's-who of education platforms that decided correctly that building and maintaining a world-class assessment layer was not their comparative advantage. What Learnosity has built over more than a decade is the curriculum tag taxonomy, the item calibration history, the accessibility compliance layer, and the response data format that allows downstream analysis. None of those components are technically spectacular on their own. Collectively, they represent a switching cost that grows every year as more of the ecosystem builds on top of them.
The AI layer makes the assessment API position more valuable, not less, because every AI tutoring and adaptive learning system needs a signal to adapt against. The dream of an AI tutor that knows exactly what a student needs next is fundamentally dependent on having high-quality, structured assessment data to learn from. A generative AI that produces customised practice problems is interesting. A generative AI that produces practice problems at calibrated difficulty levels, tagged to specific curriculum nodes, with response schemas that feed a Bayesian knowledge tracing model — that's infrastructure. The companies building adaptive AI products who think they can build the assessment layer themselves will discover that getting psychometric calibration right at scale is a decade-long project, not a sprint. The smarter approach is to treat assessment as infrastructure and build the adaptive layer on top of a well-maintained API.
For investors, the assessment API category has a specific financial profile worth understanding. Revenue comes through platform fees or per-seat API consumption, which means it scales with the adoption of the EdTech products building on top of it rather than with end-user growth directly. The sales motion is B2B2C — selling to the EdTech builders, who then sell to institutions. This creates longer sales cycles and a more concentrated customer base than a direct-to-school product. The upside is that contract values are larger, churn is very low once an EdTech product has integrated your API into its assessment flow, and the data network effects compound over time: more item responses from more student populations allows better psychometric calibration, which makes the API more valuable to the next integrator. That's a genuinely defensible compounding dynamic, which is rare in EdTech.