Define the problem well

The idea

A good problem statement says who the tool is for, what it does, and how you will know it worked.

Is that a plan?

A teammate says: Our project is to use AI to help the school. What would you ask first?

• Which AI tool should we use?
• Help whom, to do what, and how will we tell if it helped?
• How soon can we finish?

Clear, vague or too big?

Each card is a project idea. Tap a group: clear problem, too vague, or too big for one team. Read the reason, then press Next card.

Three questions

Each clear idea said who it is for, what the tool does, and how we will know it worked.

The vague ideas skipped them. The big ones could not be finished or checked by one team. The short lost-property card still passed, because it answered all three.

A good problem statement says who it is for, what the tool does, and how you will know it worked.

Teams call this writing a problem statement. It also asks whether AI is needed at all. A job done by fixed rules, like a timer, may not need it.

Notes

A good problem statement says who it is for, what the tool does, and how you will know it worked.

A quick test: could someone else tell whether your project worked? If not, make the problem sharper.

Check yourself

Zoya's class wants to help the school garden. Which project statement is the best?

A school bell must ring at 9:00 every day. Does it need AI?

Meera writes: Count cars at our school gate each hour with a camera, and check against a hand count. Which part says how we will know it worked?

• Make our school garden smarter and greener, using the best AI available.. Smarter and greener in what way? It does not say what the tool does or how anyone could check it.
• Tell the gardener each evening which plants need water, then check the soil next day. — correct. It names who (the gardener), what (water advice) and how to check (the soil next day).
• Build the best gardening AI in the world, for every garden in India.. No team could finish this or prove it. "Best in the world" has no way to be checked.
• Collect photos of every plant in the garden, one for each day.. That is a data step, and it comes later. It does not say what problem the photos will help to solve.
• No, because fixed rules are enough — correct. The job is the same every day, so a timer that follows a rule is enough. AI is for problems where the answer depends on patterns in data.
• Yes, AI can learn when to ring it. It could, but there is nothing to learn. The time never changes, so a fixed rule does the job without data or training.
• Yes, AI is a better tool than a timer. Better is not the same as needed. A timer is simpler and does this one job well. Use AI only when a problem needs learning from data.
• Count cars at our school gate. That is the what: the job the tool does.
• with a camera. That is how the tool gets its information, not how we judge the result.
• check against a hand count — correct. A hand count gives the true answer to compare with. That is how we will know the tool worked.
• each hour. That says how often the tool counts, not whether the counts are right.