Buildability
Describe what you need. AI helps you make it.
AI in education 2027
By Joseph Telfer, CEO and Founder, TopSchool
Raising the BAR in 2027
In 2027, more teachers will be able to make their own learning apps by describing what their students need. Three factors make this possible: AI can build more, building can cost less, and more educators know how to use it.
Claim Early Access Seats →Bring your teaching ideas to life. Register for PLAI™ v2 early access.
Explore the PLAI™ v2 examples →Buildability × Affordability × Readiness
BuildabilityAI helps us make it. AffordabilityWe can afford to try. ReadinessWe know what to ask.
×
BAR More teachers can build technology tools that fit their unique classrooms
The compounding effect
Easier to build.
Cheaper to try.
More teachers ready.
Imagine a teacher notices that some students cannot explain why one half equals two quarters. The teacher describes a simple app: “make a fraction wall, let students shade and compare the rows, then ask them to explain what they notice”. AI helps build a first version. The teacher checks it and tries it in a lesson.
Now put the three changes together. Buildability makes that first version easier to create. Affordability means the teacher can spend less time and money trying different versions. Readiness means the teacher has more experience with AI, can ask for useful changes and knows how to apply it in the classroom.
Each force makes the others more useful. Together, they put the ability to create innovative educational technology in more teachers’ hands.
The effect is even stronger when teachers share; creating and improving ideas together becomes easy. A colleague starts with the activity and the lessons learned, then adapts it for another class. Their changes give the next teacher a better starting point.
This is how the benefits multiply over time: when building is easier, people try more often; more tries create more experience; and sharing that experience helps the next person build something better. Teachers can then help shape the software around the real needs of their classrooms.
Affordability
Spend less time and money trying an idea.
Readiness
Give clearer instructions. Check whether the result helps.
Why more teachers could do this in 2027
What each part of BAR means.
We need tools that can build what we ask for, costs that let us try, and people who know how to guide and check the result.
B01 / 03
Buildability
AI can help make an app
and carry out tasks.
AI can write the code that makes an app work. A teacher can describe the learning goal, the examples and the support students need, then ask AI to build a first version.
You may also hear the word “agentic” or “AI Agent”. It means AI can work through several steps to carry out a request. For example, it might create a page, find a problem and change the code, rather than needing a separate instruction for every step.
It can also use other software when the systems are connected and access is allowed. A school might use this to move approved information between systems, so staff do not have to enter the same details twice.
The possibilities range from a rubric or lesson pack to an interactive game or a small school app. The teacher describes the need and checks whether the result meets it.
⌘Describe → build → check
A02 / 03
Affordability
Trying an idea can take
less time and money.
There are two costs to consider: paying to use AI, and doing the work needed to create useful software.
The first has fallen sharply for comparable AI performance. Stanford’s 2025 AI Index reported a drop from about US$20 to US$0.07 per million tokens between Q4 2022 and Q4 2024 for performance at the GPT-3.5 level. Tokens are the small pieces of text AI processes.
The second includes explaining the need, building a first version, checking it and making changes. AI can help with each step. That can make it practical to try a small app for one class, even when a larger software project would be too expensive.
Schools still need to pay for running, maintaining and supporting their tools. But when a first attempt costs less, more ideas can be explored.
Q4 2022$20
Q4 2024$0.07
Cost of AI use at comparable performance, per million tokens. Stanford AI Index 2025 ↗ R03 / 03
Readiness
More educators understand AI’s capabilities and limitations.
As teachers continue to use AI, they find out what it helps with and where it gets things wrong. That experience helps them give clearer instructions and judge the answer.
A teacher might start by asking for a quiz. Later, they may ask for questions on a specific misconception, written at the right reading level, with feedback they can check before students see it.
Training and colleagues can help teachers make that progress. The surveys below show how AI use and access to training have been changing.
If this continues into 2027, more educators could arrive with a clear idea of what they want to build, what good teaching with AI looks like and what safety standards and security protocols need to be followed.
↗Try → learn → ask more clearly
Early signs at EARCOS 2026
Teachers are already
starting to build.
FIELD NOTEEARCOS
Teaching and Learning
Conference
2026✎
Teaching and Learning
Conference
2026✎
At conferences around the world in 2026, we have seen this emerging organically. At the EARCOS Teaching and Learning Conference in Bangkok this year, I saw teachers beginning to create their own edtech and share it with each other. The change from the year before was that AI was being discussed in all the seminar rooms. In addition, some sessions were dedicated to “vibe coding”: the act of describing an idea to AI, letting it write the code, then asking it to make changes.
Teachers were trying their first mini apps and web pages in Google Sites and walking through how they could quickly try an idea, ask the person beside them for input and show each other what they had made. New ideas were being tried freely in the room.
Imagine teaching and learning conferences in 2027. A group builds a learning activity together in a workshop. Each teacher takes home a version they can change: different settings, another language, more support or a different curriculum.
The value is in both the app and the colleagues helping each other make it useful. More capable tools and more experienced teachers could make these workshops far more productive.
From early chatbots to AI that builds
The same request.
A very different experience.
With early chatbots, a teacher’s request typically led to a written answer. In this next phase, AI can take that same request and help build an activity students can use. The teacher brings the learning need. AI can do more of the work needed to turn it into something useful in the classroom.
THE SAME CLASSROOM NEEDHelp students explain why ½ = ²⁄₄
01 THE EARLY CHATBOT EXPERIENCE
A written suggestion.
Example chat response
Teacher asks
“Create an activity to help my students understand why one half equals two quarters.”
AI suggests
Draw two equal-sized rectangles. Shade half of one and two quarters of the other. Ask students to compare the shaded areas.
The teacher reads the suggestion, checks it and prepares the shapes and questions for the lesson.
↗The result is advice. The teacher turns it into an activity.
02 THE NEXT PHASE
An activity students can use.
The same request
“Create an activity to help my students understand why one half equals two quarters.”
A working Fraction Wall
Watch, then try: the wall shades matching fractions automatically. Click any tile to take over, or press Reset to start with an empty wall.
Automatic demonstration
The teacher checks the maths and instructions before students use it. Students shade the wall, compare fractions and explain why different fractions can show the same amount.
◎The result is a working activity. Students can use it in the lesson.
The request stays the same. AI can now help turn it into working software, with the teacher checking and refining the result. Students interact with the activity. The AI that helped build it recedes into the background.
What the surveys tell us
AI use is widespread.
More training is reaching teachers.
These surveys asked different groups different questions. They show how AI is being used, but their percentages cannot be compared as one exact measure of growth.
of teachers said they used AI to plan or teach lessons.
Read the survey ↗of U.S. public school teachers said they used AI for work that school year. 32% used it at least weekly for the tasks in the survey.
Read the survey ↗lower secondary teachers across OECD education systems said they used AI for work. Use varied widely between education systems.
Read the survey ↗of U.S. K-12 educators said they used AI for work. This group included teachers, teaching assistants and other classroom staff.
Read the survey ↗PD
AI training for teachers
Fewer report no training.
Education Week Research Center asked teachers about training on AI.
Q4 2024Nearly 60%reported no AI training
→
Q3 to Q4 202550%reported no AI training
Read the Education Week report ↗ What I am seeing
More trying.
More sharing.
Clearer expectations.
Over the past one to two years, I have personally heard more and more teachers talk about trying AI, attending training and helping colleagues.
They are also clearer about what they want it to do, the safety requirements they expect and what they want to keep in their own hands.
I expect more of this in 2027: teachers who have learned by trying, and who can help others get started.
From one classroom to many
One teacher’s idea.
A better starting point
for the next teacher.
Return to the fractions activity. The teacher builds it, checks it, tries it with students and changes what was confusing. When they share the activity and what they learned, another teacher can start from there.
Notice → build → try → learn → share ↺ next classroom
01 / NOTICE
Students can name a half and a quarter, but some cannot explain why two quarters equal one half. The teacher decides to try an interactive fraction wall.
One teacher may add clearer instructions. Another may translate them or adapt the activity for students who need more support. Sharing these changes can save colleagues time and give them better ideas to try. That is how the benefit can grow beyond one classroom.
What schools need to provide
Teachers can build.
Schools need to support them.
ROOM TO CREATE+CLEAR RULES
App builders such as Lovable, Replit, Base44 and Emergent let people describe what they want in everyday language. AI writes much of the code. You look at the result and ask for changes. This is often called vibe coding. It lets more people begin building without first learning to write code.
However, a general app builder like these does not automatically know your curriculum, how a school stores information, what good teaching looks like in your setting or what your learners need.
A working app also needs somewhere to run. This is called hosting. Someone must decide who can use it, what information it can collect, how that information is protected and who will fix problems.
Schools also need to check whether the activity teaches the right things, works for the students who will use it and protects children from harm. Asking each teacher to solve all of this alone creates a heavy burden.
A better approach is for the school to provide approved places to build and run apps, clear rules about information and access, and people who can help review the result. Teachers can then create different activities within those rules.
Give teachers room to create. Make the rules and support clear before an app reaches students.
This leaves plenty of space for new ideas while keeping responsibility clear. The teacher brings the learning need. The school helps make the result safe to use, maintain and share.
01What data can it use?Decide what information is allowed and who can see it.
02Where will it run?Provide secure hosting and someone to fix problems.
03Does it help the learning?Check the content, curriculum fit and support for learners.
04Is it safe for students?Check privacy and safeguarding before classroom use.
A practical start for school leaders
Give teachers time,
training and support.
Start with a small group of interested teachers and a few classroom needs they understand well. Give them time to try ideas together.
Training should cover how to describe a need, check an AI result and decide whether an activity helps students learn. Teachers should know which tools they can use, which information they can enter and who to ask for help.
Set up a clear review before students use an app. After a classroom trial, ask teachers to share the activity, what helped, what went wrong and what they would change.
Teachers need a manageable way to move from “my students need this” to “we have checked this and can try it”.
School leaders can make that possible by providing a safe “sandbox”: a place where passionate and innovative teachers can try ideas and contribute to the school’s growth.
BAR
Raise the BAR
What would you build
for your learners?
Buildability means AI can help make it. Affordability means more ideas are worth trying. Readiness means more educators know how to guide the work and check the result.
Together, these changes could put the ability to create learning apps in many more teachers’ hands. A need spotted in one lesson could lead to an activity that helps that class, then becomes useful to other teachers too.
Start with one need you know well. Find a colleague. Make a small first version, get it checked and see what students do with it. Share what you learn.
By 2027, the next useful tool for your school could begin with an idea from your classroom. Let’s give educators the support to build it.
Bring your teaching ideas to life. Register for PLAI™ v2 early access.