Building With AI: How a Small New Zealand EdTech Company Is Expanding What It Can Create

Artificial intelligence has changed the practical limits of what a very small technology company can attempt.

That does not mean AI builds everything.

It means the distance between an idea and a testable prototype is getting shorter.

At Iconic Games, we have been experimenting with AI across research, coding, debugging, planning and development workflows. More recently, that has included structured AI-assisted development inside our code repositories.

The opportunity is significant.

So are the responsibilities.

AI is leverage, not accountability

A coding assistant can propose a fix.

It cannot take legal or ethical responsibility for shipping it.

A generative model can produce educational content.

It cannot guarantee that the content is accurate, culturally appropriate or pedagogically sound.

An AI character can hold a conversation.

That does not automatically make the interaction safe for a young person.

The person or company designing the product still owns those decisions.

This is one of the most important principles in our approach to AI:

AI can increase capability. It does not transfer accountability.

What AI can do for a small development team

In practice, AI can help a small team do more iterations.

It can inspect code, suggest refactors, generate tests, draft documentation, explain unfamiliar systems and help identify where bugs might be coming from.

It can help compare approaches before engineering time is committed.

It can accelerate research and create starting points.

The phrase “starting point” is important.

The output still has to be reviewed.

In software development, an incorrect answer can break a build.

In education, an incorrect answer can misteach a learner.

That raises the standard for verification.

Responsible AI matters more in education

UNESCO’s guidance on generative AI in education recommends a human-centred and age-appropriate approach. It highlights privacy, ethical validation and pedagogical design rather than treating AI adoption as an end in itself.

UNESCO’s AI Competency Framework for Students similarly frames young people not merely as users of AI but as responsible users and co-creators who need critical judgement, ethical awareness and foundational understanding.

Those ideas align strongly with where we want Mīharo to go.

The goal is not to put AI into every interaction because AI is fashionable.

The questions should be:

What does AI improve here?

What data is genuinely required?

What should never be stored?

What happens when the model is wrong?

When should the system refuse?

When should a human remain in control?

How do we test this with the assumption that real people, including young people, will use it in unexpected ways?

AI characters need boundaries

One of the long-term possibilities for Mīharo Ahau and Mīharo Kura is conversational characters that can respond more naturally than scripted dialogue trees.

That could make learning much more interactive.

It also introduces real risks.

A conversational system may produce inappropriate information, misunderstand a learner, overstate confidence or encourage a conversation that the product was never intended to support.

So conversational freedom cannot mean unrestricted freedom.

Age-aware safeguards, topic boundaries, data minimisation, testing and clear escalation design have to be part of the system, not added after launch.

AI does not remove the need for experts

There is a temptation to treat powerful AI as a substitute for every specialist.

We do not.

A model does not replace cultural expertise.

It does not replace qualified educators.

It does not replace security review.

It does not replace accessibility testing.

It does not replace legal obligations.

It does not replace lived experience.

What it can do is help a small team reach the point where the right experts can evaluate something more concrete.

That can be incredibly valuable.

The opportunity for New Zealand founders

For a small New Zealand studio, distance and scale have always mattered.

Large studios can employ specialised teams across engineering, art, research, quality assurance, analytics and production.

Small founders cannot begin that way.

AI changes some of that economics.

It may allow more ideas to reach prototype stage before a company needs a large headcount.

That is particularly interesting for founders building in smaller towns or outside established technology networks.

The opportunity is not “one person replaces a whole industry”.

It is that more people may be able to test ambitious ideas earlier.

The standard should rise, not fall

There is a paradox here.

Because AI makes it easier to generate more, the value of judgement increases.

More code does not necessarily mean better software.

More content does not mean better learning.

More features do not mean a better product.

AI can make rubbish faster too.

So the long-term advantage may not belong to whoever generates the most.

It may belong to teams that become very good at deciding what is worth building, what needs human expertise, what should be verified and what should never be automated.

That is the direction we are trying to learn towards.

AI is part of how Iconic Games is being built.

It is not the reason Iconic Games exists.

The learner still comes first.