Classroom teachers
Break a scheme of work into concepts once, then pull practice and assessment questions from it all year.
A workshop for teachers human-guided AI
A workshop for your subject.
You bring the material. You keep the questions.
KIE is a workshop, not a shop. Nothing is pre-made and nothing is sold to you off a shelf. You bring your own course material, break it into concepts, connect them, and generate questions from the structure you approved. AI is the tool on the bench — you're the one making the thing.
Your material, your structure
Built from your source, not a stock library
Questions with a source
Every question traces back to an approved concept
Built once, runs for years
The structure keeps producing after the work is done
KIE specialises in three areas of question intelligence.
01
Reverse engineering
Understand how a question is built
Start with an existing or past-paper question. KIE breaks it back into the concepts, relationships, information, reasoning path and thinking demand behind it. The teacher can then rebuild the same knowledge into a new question without simply copying the original.
Explore reverse engineering → 02 Multi-concept engineeringBuild questions from connected concepts
Combine one, two or more approved concepts only when their relationship matters. KIE lets the teacher control which concepts are joined, how they interact and how much conceptual load the learner must handle in a single question.
Explore multi-concept questions → 03 Higher-order thinkingChoose the thinking the question demands
Move beyond easy, medium and hard. Build questions that require learners to understand, apply, analyse, evaluate, create or transfer knowledge. Difficulty comes from the reasoning architecture, not simply from longer wording or larger calculations.
Explore higher-order thinking →KIE — Knowledge Intelligence Engine. Three words, and each one is doing a job.
Knowledge
Not files and not text. The subject broken into concepts, the links between them, worked explanations, examples and the mistakes learners actually make — held in a structure you can open and edit.
Intelligence
Reasoning about that structure: which concept a question really tests, what depth it demands, which relationship it depends on, and what a correct answer has to show.
Engine
Something you build once and run repeatedly. Engineer a concept properly and it keeps producing questions, explanations and practice for as long as you teach the subject.
Put together: KIE engineers knowledge rather than generating text. A chatbot writes a plausible question and forgets it. A shop sells you someone else's. An engine builds the knowledge first, keeps it, and derives the question from it — which is why the work you put in on Monday is still working for you next term.
Who it's for
KIE is for teachers building questions for their students
It isn't a study chatbot and students don't use it to generate anything. A teacher engineers the knowledge and approves the questions; students get a practice area containing only what that teacher published.
The teacher's side
Where the work happens
- Bring in the course material
- Break it into concepts and connect them
- Write the explanation, examples and misconceptions
- Generate questions, then review every draft
- Publish what's good, bin what isn't
The student's side
What they actually see
- Approved questions and nothing else
- The explanation their teacher wrote
- Worked examples and another way of explaining it
- A confidence rating after each attempt
- Revision that comes back when it's due
Which suits three kinds of teaching in particular:
Tutors and exam coaches
Take a past paper question, see exactly what it tests, and produce a new one at the same standard in a different setting.
Course teams
Agree the concept map as a department so everyone teaches and tests against the same approved structure.
How it works
Six steps, in the order you'd actually do them
The first topic takes real work — this is a workshop, not a vending machine. What you get for it is a structure that keeps producing questions long after the setup is done, and each step leaves something you can check before the next one starts.
Bring in the material
Add books, notes and handouts, and set the subject boundary so unrelated material stays out.
Split it into concepts
Large subjects become topics, subtopics and concepts small enough to teach, check and reuse on their own.
Connect them
Say how concepts relate: what depends on what, what explains what, what should be compared with what.
Write the explanation
Add the meaning, the first principle, a worked example, a real-world use and the mistake learners usually make.
Generate questions
Choose the concepts, the type of thinking and the context. KIE drafts the questions; you keep the ones that work.
Publish to learners
Approved questions and explanations go to the practice area, where learners work through them at their own pace.
The engine
What goes in, what comes out, and who decides
This is the knowledge engineering loop in full. Three things feed the engine: your approved subject content, your rules about what counts as correct, and a language model that drafts and analyses. Nothing leaves the engine without passing you first.
The model never publishes
Every draft stops at your review. That one gate is the difference between a question bank you trust and one you have to re-check.
Answered once, reused after
Approved material is stored and reused for later questions, so the same ground isn't paid for twice and the wording stays consistent.
Only generate costs anything
Saving, loading and approving are free. The model is called when you ask it to draft or analyse, and not otherwise.
The structure
Every question knows where it came from
Most tools generate a question and forget it. KIE keeps six layers connected, so you can always open a question and see which concept it tests, which relationship it depends on, and which part of your source material it came from.
Subject
Sets the boundary. Nothing outside it can leak into a question.
Topic
A major area of the subject, roughly the size of a unit.
Subtopic
A narrower area inside the topic, closer to a single lesson.
Concept
The principle, process or idea a learner has to understand.
Relationship
How one concept depends on, explains or contrasts with another.
Question
Tests recall, application, reasoning, comparison or transfer.
The question engine
More than the same question, reworded
A reworded question tests the same thing twice. KIE changes what the question asks the learner to do — and it works in both directions: from a concept to a question, and from a question back to the concepts behind it.
Keep it focused
When the learning objective is clear, one concept is enough. No padding, no unrelated data to wade through.
- Definition and first principle
- Straight application
- Checking a common misconception
Test the connection, not just the parts
Combine concepts when the link between them is the point. Two is the normal case. Three at once only where the question genuinely needs all three, otherwise you're testing stamina rather than understanding.
- Reasoning across a relationship
- Comparison and dependency
- Integrated problem solving
Start from a question you already have
Paste in a past paper question. KIE works out what it really tests, what reasoning it needs, and what information in it is doing nothing.
- Which concepts it draws on
- Which relationships it depends on
- Its dimension and depth
- Missing or unused information
Change the setting, keep the principle
The same reasoning, moved to a different organisation, sector, role or constraint. This is where you find out whether it was understood or memorised.
- Real-world application
- Familiar situations first
- Then unfamiliar ones
Multi-concept
Real problems rarely sit inside one concept
A learner can know NPV and know what a discount rate is, and still not see how one moves the other. That gap only shows up when a question makes them use both at once. KIE builds those questions from the relationship you have already approved, so the connection being tested is one you chose.
2
Two concepts is the normal case
Enough to test a relationship without burying it in detail.
3
Three only when all three are needed
For genuinely integrated questions. Past that you're testing stamina, not understanding.
1
One subject boundary, always
Concepts are only combined inside the same subject, so nothing unrelated leaks in.
✓
Approved concepts only
If you haven't approved both concepts and the link between them, KIE won't combine them.
How a combined question is built
Concept A
Net present value
+
Concept B
Discount rate
Approved relationship
A higher discount rate cuts the present value of later cash flows more than earlier ones.
Generated question
The board wants the appraisal redone at 12% rather than 8%. Explain which of the two projects now ranks higher, and why the timing of the cash flows matters more here than the total.
The reasoning path
Subject →
Concept A +
Concept B →
Relationship →
Higher-order question →
Reverse engineer →
New context
Once the question exists, the same structure runs backwards: KIE reverse engineers the reasoning and writes a fresh question on it for a different company, sector or set of constraints.
Build the knowledge before you test it
KIE develops the concept itself first — what it means, the principle underneath it, how the process runs, what it connects to, where it is used and where learners usually go wrong. The question engine then has something solid to work from instead of guessing.
Meaning
First principle
Process
Relationship
Application
Misconception
Start with a question and find the knowledge behind it
Paste in a past paper question. KIE works out which concepts it draws on, which relationships it depends on, its dimension and depth, the reasoning pattern it expects, and which numbers in it are doing nothing at all. Then it writes a new question that tests the same knowledge somewhere else.
Question →
Concepts →
Relationships →
Reasoning →
New context
Useful two ways round: to check that a question you wrote tests what you think it does, and to turn one past paper question into a set.
Higher order thinking
Difficulty isn't just easy, medium and hard
Two questions can carry the same marks and still ask for completely different work. In KIE you set the level of thinking a question demands — understanding, application, analysis, evaluation, creation or reasoning across several concepts at once.
Build the mental model
Meaning, first principles and why the concept works, before anything is manipulated.
Use it in context
Calculations, decisions and practical situations, including examples not seen in class.
Break the problem apart
Spot the concepts, assumptions, relevant data and the reasoning path a solution needs.
Compare and justify
Weigh alternatives, explain trade-offs and say when a method stops being suitable.
Construct something new
Design a solution, a scenario or a question by combining approved concepts.
Move it somewhere unfamiliar
Same principle, new setting. The clearest test of whether learning has stuck, and where multi-concept questions do most of their work.
Experience-based learning
Start with something the learner has already seen
A hard concept gets easier when it is attached to a familiar place, object or decision. KIE can open with that experience and walk outwards from it, one step at a time.
01
Experience
A place, event or situation the learner already knows.
02
Observation
What is happening there, and what decisions are being made.
03
Concept
The principle underneath the thing they just described.
04
Reasoning
Explain, compare, calculate or justify using that principle.
05
Transfer
Apply the same reasoning somewhere they have never seen.
Teacher control
AI widens the search. You decide what counts as correct.
KIE uses AI to suggest and analyse, never to publish. Knowledge, relationships and questions all sit in a review queue until you approve them — and once approved, they are reused instead of regenerated, so the same material doesn't drift between lessons.
AI suggests concepts, relationships, explanations and questions.
You check the accuracy, the terminology and whether it's worth teaching.
Approved material becomes part of the trusted structure learners see.
It's stored and reused for later questions, so nothing is generated twice.
In development
Next: the answer should decide what comes next
A score on its own says very little. The learning layer we're building reads a learner's answers as evidence — what they understand, which relationships are weak, which misconception keeps returning — and uses that to choose the next explanation or question. It isn't finished yet, and this page will say so until it is.
Questions
Before you sign up
Does KIE replace the teacher?
No. It drafts, and you decide. Nothing reaches a learner until you have read it and approved it, and you can rewrite anything before you do.
Which subjects does it work with?
Any subject you can describe as concepts and relationships. Boards have been built for maths, chemistry and financial management so far; the structure itself is subject-neutral.
How do you stop it inventing facts?
Questions are generated from concepts you have already approved, inside a subject boundary. Where a real-world example is used, KIE is told not to present it as current fact unless it appears in your source material.
Does every click cost an AI call?
No. Saving, loading and approving are free. Only the generate and analyse actions call the model, and approved answers are stored and reused rather than regenerated.
How long before I have something usable?
Reverse engineering is the quickest way in — paste a past paper question and you get the concepts behind it plus new versions in minutes, and the concepts are saved as you go. Building a topic properly from scratch takes a few hours. That's the trade: more work at the start, a structure that keeps producing afterwards.
Can I try it before committing?
Email hello@kieindia.online with your subject and level, and we'll set you up with a workspace and a sample board.
Bring your subject. Leave with something that keeps working.
Your judgement, AI on the bench, and a structure you can actually see.