AI Maths Tutor Review: Deputy Headteacher Neil Almond’s Independent Assessment

Neil Almond is a respected deputy headteacher and primary school expert working with the STEP ahead teaching school hub. He is also an experienced maths teacher who is now championing the strategic use of AI in schools. 

We asked Neil to independently trial Skye, our spoken AI maths tutor, and share his thoughts on the pedagogy, usefulness, and real-world application.

Neil tested Skye’s SATs Revision Programme, over several weeks in March 2025. Since then, not only have we implemented some of his recommendations, but we’ve also added standard primary maths programmes for Years 4, 5 and 6, and our series of GCSE maths programmes. 

Here’s Neil’s independent assessment of the AI tutor Skye:

Over the past few weeks, I’ve had the opportunity to test out Skye, Third Space Learning’s new AI maths tutor, designed to support primary maths learners – and GCSE students from September. I was placed on their SATs Revision Programme — a timely addition to any Year 6 teacher’s toolkit. Having spent time exploring the platform, here are my reflections on the usability, pedagogy, and potential of this early iteration.

Interface and usability

Skye’s interface is clean and uncluttered, a refreshing contrast to the often overstimulating design of other edtech tools. 

Cognitive load has clearly been considered: colour, cursor movement, and dynamic input fields direct attention purposefully. The answer box appears only when relevant, encouraging pupils to attend to each step rather than race ahead.

The AI maths tutoring onboarding experience lasts around 7–8 minutes and is deliberately paced. Pupils are introduced to the voice input system — which works surprisingly well — and guided through how to interact with the platform. It builds rapport with return questions and conversational responses. Importantly, it avoids the trap of mimicking a human teacher — a wise decision.

I was asked to complete an initial assessment, though being placed directly on the SATs revision course meant I wasn’t best positioned to judge its accuracy or impact on what the system offered.

Booking and scheduling

Throughout my period of testing, an unusual design choice was made that, despite being entirely AI-driven, Skye sessions must still be booked in advance. This could only be done Monday-Friday and only between set hours. 

However, since my testing period has come to an end, pupils are now able to launch a Skye session on demand on any day of the week they choose. 

This is a win-win. Teachers can set sessions for homework, or parents can encourage its use at home.

Pedagogy and pupil experience

Skye follows a structured instructional sequence: one worked example, one guided “we do”, followed by two independent “you do” tasks. This mirrors effective classroom practice and will feel familiar to pupils.

However, if this remains the sole structure, some learners may find it monotonous. Including more open-ended tasks or classic mathematical problems could broaden the experience and support both motivation and mathematical thinking.

The platform includes motivational elements such as badges and positive reinforcement, but these are used sparingly and support rather than distract from learning.

At present, Skye’s adaptivity appears to be rule-based. While it responds to correct and incorrect answers, it doesn’t yet tailor scaffolding or progression dynamically based on pupil performance. Future iterations could benefit from more nuanced adaptivity to maximise learning impact.

I am told that such behaviour is coming, but I have not seen it in practice. However, the rate of improvements from Third Space since I first gained access to Skye leads me to believe that they will make good on these improvements.

Feedback and misconceptions

If a pupil answers incorrectly two or three times, Skye reveals the correct answer. In many cases, this is useful. However, it sometimes moves on too quickly without addressing the underlying misconception. A moment to pause or revisit the error could make these sessions more instructionally effective.

There were also instances where feedback didn’t fully align with the learning objective, for example, referencing column addition when the focus was on adjusting. Tighter coupling of feedback to task intent would strengthen the learning experience.

Safeguarding and support

Skye incorporates sensible safeguarding protocols. If a pupil expresses concerning language, such as “I feel like I want to hurt myself,” the system issues a stock response advising the pupil to speak to a trusted adult and flags the concern with school staff. 

All sessions are recorded, and any alerts are followed up appropriately. It’s a thoughtful integration that reflects a strong commitment to pupil wellbeing.

Final reflections

Skye is still in its early stages, but there’s already much to commend. The interface is intuitive, the instructional design is grounded in sound pedagogy, and the development team has been impressively responsive to feedback. 

For schools seeking a flexible, low-effort way to extend support, particularly where teacher capacity is stretched,  I’d recommend exploring Skye’s AI tutoring. It is a platform worth watching and, increasingly, using.

You can also read EdTech Ninja Jodie Lopez’s AI tutoring review on how Skye provides tutoring to pupils who need it where tight budgets mean support might be lacking.