AI can produce a page of text in seconds. The more difficult question is what that makes possible for the educator and child.
The argument I developed for the Montessori Australia National Conference starts with the relationship. I want us to judge AI by whether it helps educators give children their attention, use their professional judgement and make thoughtful choices about learning.
That was the question behind my session at the 2026 conference in Brisbane. The organiser's account of the session describes a conversation about purpose, leadership and culture. This article develops the human role at the centre of that argument.
Begin with what you want to protect
Before listing the tasks a tool could complete, I would ask a team what they want children to experience.
An adult who notices. Time to explore an idea. Space to attempt something before an answer arrives. Someone who can recognise when to help, when to wait and when a child is inviting a response.
Those commitments give us a basis for evaluating technology. A new feature can sound useful in isolation. Its value becomes clearer when we ask what happens in the room because of it.
Does the educator gain time and attention for the child? Does the tool help them think about what they have noticed? Can they see and question how an output was produced? Or has a new task, screen or checking burden been added to their day?
These questions place the purpose of the work at the beginning of the decision.
Protect the educator’s judgement
There is a difference between helping an educator express what they have seen and deciding what a child’s experience means for them.
I want technology to support the educator’s thinking. That means starting from something the educator actually noticed, keeping that observation visible and giving them room to interpret it.
Consider an illustrative example. A child returns to a construction several times, changes its base and invites another child to join. An AI-generated account might read fluently. The educator still needs to consider what happened, what the child was trying to do, what remains uncertain and whether another perspective would change their interpretation.
The quality of that thinking matters. We should be able to ask who supplied the evidence, which words describe an observation and which words are an interpretation. A confident sentence should never make those distinctions disappear.
For me, useful assistance leaves the educator more able to explain their choices.
Leave room for discovery
In preparing the session, I kept returning to another question: when does an answer help, and when does it arrive too soon?
A child trying, revising and trying again is doing something valuable. My concern is that an always-available answer can tempt adults to close a question before the child has had time to explore it.
That does not produce a universal rule against answering questions or using technology. It asks us to make a pedagogical choice. What is this child exploring? What would a response open up? What might waiting make possible?
A tool should fit those decisions. It should not quietly make them for us.
Decide where the time goes
Reducing repetitive work is an understandable goal. I think the next question deserves equal attention: what do we intend to do with the capacity we create?
If every saved minute becomes an expectation to produce more material, the educator may finish with the same pressure and a longer record.
A team can choose a different purpose. It might use that capacity to discuss an observation with a colleague, listen more carefully to a family or spend a less interrupted period with children. These are possible choices, not outcomes a tool can guarantee.
Leaders can make the intention explicit before introducing a new workflow. They can then ask educators whether the experience matches it.
Try a small, purposeful evaluation
A useful starting point is one task and one clear intention.
Choose a piece of work the team wants help with. Describe the current process, including the time spent checking and correcting it. Agree what must remain a human responsibility. Try the proposed assistance with appropriate information safeguards, then review what happened.
The review can be simple:
- Did the output stay faithful to the educator’s evidence?
- Could the educator identify and correct assumptions?
- Did the process support their thinking?
- Did it create the time or attention the team intended?
Use the answers to decide whether the workflow deserves a place in your service.
Return to why you chose the work
One reflection in my prepared session asked educators to remember why they first chose to work with children.
That question remains useful when a technology decision becomes crowded with features. What part of that original purpose do we want to make easier to live out? What responsibility do we want to hold onto?
My hope for AI in early childhood is that it helps us create the conditions for more thoughtful, present human work. We will need to make deliberate choices for that to happen.
For the wider operator discussion, read How to win in the age of AI in ECE. With your own team, begin by choosing one human responsibility you want any new tool to protect.
