Before
A child practising at home needs material that matches what school actually teaches, and a parent cannot see where that child is struggling until a report arrives at the end of term. A chatbot is the obvious build and the wrong one: it is fluent, occasionally wrong, and the user is a child who has no way to tell the difference.
What we built
- Built Learn, Practise and Quiz for each curriculum topic, so a child moves through a topic in a structure rather than chatting with a model
- Made the marking deterministic: a symbolic maths solver decides whether an answer is right, and a grammar engine checks written work. The model never gets a vote on correctness
- Built a parent view of progress and weak areas, so involvement does not depend on a child volunteering what went badly
- Wrote the child-safety and education boundaries into the product and published them: supervised, age-appropriate use with parental oversight as part of the design
What changed
- A child gets curriculum-linked explanations, practice and feedback in plain language, with encouragement that does not punish a wrong answer
- A parent can see progress and weak areas without waiting for a school report
- Marking cannot drift with a model's confidence, because no model is marking
Where a person stays in charge
- No claim about grades or academic outcomes. The product supports learning and does not replace teachers, schools or parents.
- No model decides whether an answer is right. That is arithmetic, and it is the same rule that stops an invented figure reaching a board memo in our client work.
- Designed for children in supervised, age-appropriate contexts, with parental oversight built into the product rather than added to the terms.
How it is measured
Marking is checked against the solver, not a model, so correctness is testable rather than judged, and the checks run on every change. We make no claim about grades or academic outcomes and do not publish a figure for them.
What stays confidential
Nothing is anonymised here, because the product is ours and you can open it. We do not publish user numbers, and we will not publish an outcome claim we cannot measure.
Same shape as your process?
Tell us what the equivalent task costs you today and we will tell you what this shape would take to build for you, what it should return, and where to start. Thirty minutes, no slides.
Services this came from
Make the next AI project one the business can measure.
Thirty minutes with the founder, no slides. You will know what the right solution looks like, what it would take to build, what it should return, and which part to start with.
Replies come from a named person in Dubai within one working day.