Perspective 4Design + AI Mindset · Education & Careers

When AI Can Answer, School Must Teach the Questions

Indian schooling still rewards memory and obedience. In a world where AI can produce an answer in seconds, the scarce skills are knowing what to ask, judging what comes back and learning from failure.

The topper who could not start

The story below is illustrative. It is not about a real person.

A student scores in the top tier of her board exams. In her first college project, she is asked a simple question: what problem in your city would you most like to solve, and why? She freezes. For twelve years, every question came with a single right answer, a marking scheme and a deadline. Nobody asked her to choose the question.

She is not short of ability. She was trained for a world that is disappearing.

What the national survey shows

In 2024, India assessed over 21 lakh students in Classes 3, 6 and 9 across 781 districts in the PARAKH Rashtriya Sarvekshan. Class 9 students averaged 54% in language, 40% in science and 37% in mathematics. Scores fell from Class 3 to Class 9 in both language and maths.

Class 9 scores fall furthest in maths

Average scores of Class 9 students, PARAKH Rashtriya Sarvekshan 2024

Language54%
Science40%
Mathematics37%

Bars start at zero.

Class 9 average scores in the national survey. Source: PARAKH Rashtriya Sarvekshan 2024, via The Federal.

Source: PARAKH Rashtriya Sarvekshan 2024, via The Federal. Better India By Design. See article sources →

Assessment averages, not the percentage of pupils passing. Different subjects, same Class 9 assessment.

View original chart →

At the same time, employers expect 39% of workers' core skills to change by 2030, and rank analytical thinking, creative thinking and curiosity among the skills that matter most.

More is in place than most people know

  • A policy that asks for it. The National Education Policy 2020 calls for critical thinking, experiential learning and less rote memorisation.
  • National AI infrastructure. The IndiaAI Mission is building compute, datasets and skilling for AI.
  • Skilling at scale. Programmes such as FutureSkills Prime teach AI and data to lakhs of learners.

The intent exists. What is missing is a classroom design that turns AI from a shortcut into a thinking partner.

Why the old model is stuck

  • Assessment rewards answers, not reasoning. Students learn what earns marks.
  • Failure is punished. Without safety to be wrong, nobody experiments.
  • AI is taught as a tool, not as something to question.
  • Learning is separated from real problems. Few students ever work on something that matters outside the classroom.

The redesign: a six-step blueprint

A design to test. Not a claim of achieved outcomes.

  1. Curiosity first.

    Start units with a real question from students' own lives, not a definition to memorise.

  2. Problem framing.

    Before solving, students write down whose problem it is and what a good outcome looks like.

  3. AI fluency.

    Students use AI to explore, then explain what it got wrong.

  4. Evidence judgement.

    Every claim must be checked against a source. Spotting a confident error earns credit.

  5. Human and AI creation.

    Students build something, with AI as a collaborator and themselves as the authors.

  6. Public proof of work.

    Work is shown to people outside the class, such as parents, a local office or a business.

Throughout, teachers become guides rather than lecturers, and mistakes are treated as data.

Where AI helps, and where people decide

AI can be a personal tutor, a debate partner that pushes back, a generator of practice problems and a simulation engine for "what if" questions. Used well, it gives every student a patient helper.

Teachers remain responsible for judgement, ethics and collaboration. Assessments that decide a student's future must stay with people, with a way to appeal.

Start with one cohort

One college cohort of 150 learners, or a group of schools, can run one semester of cross-subject challenges built on the six steps. Our planning estimate is ₹20 to 30 lakh.

The measures that matter are the quality of problem framing, how often students verify sources, projects completed, and whether skills transfer to a new subject.

What each of us can do

Who can act and the next step
The Ministry of Education, NCERT and state boardsadd marks for reasoning and source-checking, not only for final answers.
School leadersrun one real-world project per term with a local partner.
Teachersask students to find what AI got wrong.
Parentspraise good questions as much as good marks.
Studentskeep a record of problems you framed and what you learned when you were wrong.

Teach the questions

India's children are not short of intelligence. They are short of practice at the things machines cannot do for them. When AI can answer, school must teach the questions.

Watch the full episode →

Redesigning Indian Minds: A Six-Step Blueprint for the AI Era

Your turn: teachers and parents, what is one skill children need that schools do not teach yet? Tell us in the comments on the video, and subscribe to Better India By Design for the next redesign.

Watch on YouTube →

Sources and notes

Figures reflect the periods, populations and definitions in the cited material. They are not live totals.

  1. 01
    The Federal, "What PARAKH 2024 survey tells us about learning gaps among Indian school students", 15 July 2025. Link
  2. 02
    FM Magazine, on the World Economic Forum's Future of Jobs Report 2025, February 2025. Link
  3. 03
    IndiaAI Mission. Link
  4. 04
    FutureSkills Prime. Link

Pilot costs are Better India By Design planning estimates.

Researched with AI assistance. Facts checked against the cited sources and reviewed by the BIBD Editorial Team.

BIBD proposals are designs to test, not claims of achieved outcomes. Pilot costs are planning estimates unless a source states otherwise.

Website edition prepared from the supplied Perspective on 31 October 2026. Original document →

Continue reading

Another system. A connected question.