The short answer
Genuine personalised learning needs three things most products lack: a model of what each student actually knows, content generated or selected against that model, and feedback that explains rather than grades. AI makes the second and third feasible at scale. The first is still the hard part, and it is a data problem more than a model problem.
What Exec9 learned building ThinkFirst
ThinkFirst is a modern EdTech web application designed to transform grammar learning in schools through AI-powered content and personalised student experiences, built with the MERN stack and ChatGPT.
- Generated content must be checked against a curriculum, not just produced. Fluent and wrong is worse than nothing in education.
- Feedback that says why an answer is wrong changes outcomes. A score does not.
- Teachers need visibility into what the system decided about a student, or they cannot trust it in a classroom.
Where AI genuinely fits in education
Content generation at scale
- Producing exercise variants per student, which no curriculum author can do by hand.
Explanatory feedback
- Turning a wrong answer into a targeted explanation, instantly, at any hour.
Progress modelling
- Detecting which concept a student is actually missing rather than which question they failed.