Switch course ▾
Session 1. Where AI Adds Value, and Where It Does Not
Objective. Lead a group in building a one-page value map for your school or cluster: name real problems, list possible AI uses in teaching, assessment, administration and communication, and sort each use into start, pilot with conditions or do not do, with a reason.
Most AI mistakes in schools are not technical. They happen when a tool arrives before anyone has named the problem. A leader who starts from problems saves money, protects colleagues' time and keeps trust. This lesson goes beyond the key ideas: it shows how to find problems worth solving, how to describe them so that progress can be seen, and how to handle the conversation when a senior colleague, a parent or a vendor already has a favourite solution.
Key ideas
Strategy starts from problems
- A strategy that begins with a tool asks 'where can we use it?'. A strategy that begins with a problem asks 'what would fix this, and is AI a good means?'. The second question protects money and trust.
- Write each problem as one sentence: what, for whom, how you will notice it improving. For example: 'Grade 9 writing feedback reaches pupils after three weeks; we will notice when it takes one week.'
- Course 11.7 gave you a policy and a pilot for one tool. Here you decide which problems deserve AI at all, across the whole school or cluster.
Four areas to scan
- Teaching: drafting texts at several reading levels, quiz and exercise drafts, ideas for explanations. The teacher checks and adapts everything.
- Assessment and administration: AI can draft items, summarise long circulars or propose timetable options. It must not decide marks, discipline or who gets support.
- Communication: draft notices and translations for parents, checked by a bilingual colleague, with no names or private details typed into a tool.
The five-question value filter
- 1 Is the problem real? 2 Is AI better than what we do now in time or quality? 3 Can a teacher check the output cheaply? 4 Is the risk low and the step reversible? 5 Does it work for every pupil and on paper?
- Five yes answers: start. One or two no answers that can be fixed: pilot with conditions. A no on question 4 about people or children: do not do.
- Write the reason next to every placement. A zone without a reason cannot be defended to a colleague or a board.
Where AI should not lead
- Decisions about people: marks, discipline, special-needs placement, wellbeing concerns. UNESCO guidance keeps teachers and human agency at the centre.
- Learning that depends on struggle: if pupils never write a first draft themselves, they do not practise the skill you want them to learn.
- Anything that collects data on children without a clear need, such as cameras that score attention. Session 4 returns to this.
Three zones of a value map
Start
Five yes answers
- Draft reading texts, teacher checks
- Summarise a circular, head checks
- Quiz draft, teacher solves each item
Pilot with conditions
Fixable no answers
- Parent notices with a bilingual checker
- Pupil questions after own first draft
- Owner and date for each condition
Do not do
Decides about people or children's data
- AI sets marks
- Lists of predicted failures sent home
- Cameras scoring attention
Finding problems worth solving
Ask three groups. Teachers: where do you lose hours each week? Pupils: what is hard to get help with? Parents: what do you wish you heard from us earlier? Then check the answers against records you already have: dates on feedback, notice dates, timetable clashes, the list of materials teachers make themselves.
Rank the problems by size, how often they occur and how fixable they are. Be ready to find that the best fix is not AI at all: a shared template, a better timetable or a short training. A strategy that openly says 'this one needs no AI' is more believable than one that promises AI everywhere. Write the top three in one sentence each before you look at any product.
Reading the zones with care
A zone is a decision with a reason, not a label for ever. START uses still need a named teacher who checks. PILOT uses carry conditions with an owner and a date, for example 'a bilingual colleague checks every notice'. DO NOT DO items are reviewed each term, because tools and rules change.
Look for the question that decided the placement. If question 4 fails because a decision about people or data on children is involved, no condition fixes it. If two colleagues disagree, write both reasons down. New colleagues who read the reasons will not reopen the same argument each year. Put the date of the next review at the top of the sheet.
Examples from the classroom
Counting before choosing
Scene. Dilnoza, a methodist in a Samarkand school, asks the Grade 5 and 6 team where their hours go. Most say 'texts at different reading levels'. She asks six teachers to note minutes for one week. The total is about six hours for the team. That number, not a conference slide, becomes the problem statement for the board and shows what a tool would have to improve.
Insight. A small count turns a feeling into a problem that can be tested.
The fix that needed no AI
Scene. In Fergana the head wants AI-written report comments. Jasur runs the filter: the tool would need pupils' names and marks, so question 4 fails. The real problem is that teachers write the same 20 phrases again each term. Jasur proposes a printed comment bank by subject, written once by teachers. Preparation falls and no data leaves the school.
Insight. A clear problem often has a simpler fix than the tool first proposed.
Writing the reason down
Scene. A vendor offers Aziz in Bukhara a camera that scores each pupil's attention. The team places it in DO NOT DO and writes two sentences: it collects data on children and decides nothing a teacher cannot see. Two terms later a new deputy head asks about the camera. Aziz hands over the sheet and the discussion takes five minutes, not an hour.
Insight. Recorded reasons make decisions durable and save time for the next leader.
Common pitfalls
- Starting with the tool or the conference demonstration instead of the problem.
- Treating a zone as permanent and never reviewing it.
- Writing zones without reasons, so that nobody can defend them to a board.
- Assuming every problem needs AI, and missing a simpler fix.
Build the Value Map
- Open the Activity Kit: a SAMPLE cluster (Chinor Cluster, five schools, 2,900 pupils, 180 teachers) with three problems and twelve proposed AI uses (5 min).
- Pairs (15 min): apply the five filter questions to each use, marking yes or no in the table.
- Pairs (15 min): place each use in START, PILOT WITH CONDITIONS or DO NOT DO, and write the reason in one sentence. Link every START use to one of the three problems.
- Groups of four (10 min): compare maps. Where two pairs disagree, the pair that placed the use in the safer zone explains its reason, and the group decides.
- Part B (12 min): write three problem statements from your own school using the template, then choose the one you would address first and say why.
- Alone (3 min): write your leadership task from the Kit.
Your facilitator will guide you through this activity in the session.
Questions to discuss
- Which use was hardest to place, and which filter question decided it?
- Did any START use fail to link to a real problem? What does that tell you about buying by enthusiasm?
- Which DO NOT DO items would be most tempting for a busy leader, and why?
Reflect
- Which problem in your school would you most like to fix, and is AI the best means?
- Where are you under pressure to adopt something before you can name the problem?
- Which colleague would challenge your value map best?
Try it tomorrow
- Ask three colleagues, 'where do you lose the most time each week?' and count the answers.
- Choose one problem and write it in the template: what, for whom, how we will notice improvement.
- Take one tool a vendor has offered you and list which of your problems it solves.
Further reading and links
Name the problem, run the five-question filter, give a reason for every zone, and keep decisions about people with people.