Third Space Learning Alternative: Skye vs Dilr Academy

Third Space Learning is a well-established name in one-to-one maths support for UK schools. It has worked with schools since 2013, and it now delivers its tutoring through Skye, a spoken AI maths tutor that talks pupils through problems. So when a trust or college searches for a Third Space Learning alternative, the real question is rarely "who else tutors maths". It is whether the job they need done is a maths intervention at all.

This guide is written for a multi-academy trust education lead or a college curriculum lead who is weighing Skye against a different kind of AI tutor. It compares Skye with Dilr Academy on the mechanics that decide a shortlist: what each tutor teaches, how it teaches, what evidence sits behind it, and who carries the safeguarding duty. It concedes plainly where Third Space Learning is the better choice. It does not repeat the full DfE standards walk-through, which lives in our AI tutor safeguarding checklist, or the multi-vendor buying process, which lives in our AI teacher buyer's guide.

This guide is shipped by the team behind Dilr Academy, an AI tutor that builds interactive, multilingual courses on demand with live diagrams and Socratic questioning. Or see DATS, the five-stage consulting system we use to place AI inside an organisation's existing operations.

What is Third Space Learning, and what does Skye do?

Third Space Learning is a UK education company, founded in 2013, that sells one-to-one maths intervention to primary and secondary schools and multi-academy trusts. It now delivers its tutoring through Skye, a spoken AI maths tutor trained on KS2 and GCSE problems. Skye talks a pupil through each step of a lesson written by former teachers, and the school chooses which pupils, topics and schedule it covers.

The company's own facts page for AI assistants describes the shift directly: after years of online tutoring by human tutors, it now delivers all tutoring through Skye. Its AI maths tutor page says the first pupils and teachers trialled Skye in the summer term of 2024, and that Skye is trained to teach KS2 and GCSE maths problems using modelling, scaffolding, real-life contexts and verbal reasoning. The defining design choice is voice. Unlike a text chatbot, Skye speaks, listens and guides a pupil through a problem, so the pupil reasons out loud.

The second design choice matters just as much for a buyer. Skye does not write its own lessons. The company's AI tutoring and safeguarding page states that Skye does not generate lessons and cannot display anything on screen other than lessons created by its academic team of former teachers. The how it works page sets out eight steps: the school adds pupils, sets a schedule, selects programmes and orders lessons; Skye tutors and adapts the pathway and pace with in-session assessments; and staff can launch ad-hoc sessions and monitor progress reports. Most of the decisions sit with school staff, and the teaching itself sits with Skye.

Read together, those pages describe a tightly scoped product: a maths intervention tool for school-age pupils, with a fixed bank of teacher-written lessons and a voice tutor that delivers them. That scope is a strength for the job it was built for, and it is also the reason some buyers look elsewhere.

Why do schools look for a Third Space Learning alternative?

Schools and colleges look for a Third Space Learning alternative when the need sits outside its scope: a subject other than maths, a learner older than GCSE, a course the lesson bank does not contain, or teaching in a language its pages do not describe. Third Space Learning is built for school maths intervention, so a buyer whose problem is broader usually needs a different kind of tutor rather than a cheaper copy of Skye.

In our reading of the market, the search for an alternative usually starts from one of four situations.

1. The subject is not maths. Skye is a maths tutor. A trust that wants one-to-one support in science, computing, a modern foreign language or a vocational subject cannot get it from Skye, however good the maths is.
2. The learner is past GCSE. Skye is trained on KS2 and GCSE problems. A sixth-form college, an FE college running adult numeracy, or a university foundation year has learners outside that range.
3. The course does not exist yet. Because Skye only teaches lessons the company has written, a school cannot ask it to teach a topic outside that bank. A college building a new short course has no lesson bank to draw on.
4. The course must run in another language. A college with a large multilingual intake may want tutoring in the learner's first language, including right-to-left scripts, which none of the Third Space Learning pages we reviewed describe.

If none of these four applies, and the need is a defined maths intervention cohort from Year 3 to GCSE, the honest answer is that you may not need an alternative at all. We say more about that below. The point of this section is that "alternative" is a statement about scope before it is a statement about vendors, which is the same lesson our Sparx Learning comparison drew for homework platforms.

How does Skye compare with Dilr Academy as a tutor?

Skye and Dilr Academy are both one-to-one AI tutors that use voice, but they are built for different jobs. Skye delivers a fixed bank of teacher-written KS2 and GCSE maths lessons to school pupils. Dilr Academy builds structured, multilingual courses on demand across subjects, teaches them with Socratic questioning in voice or text, and tracks mastery with knowledge tracing. The choice depends on whether you need a maths intervention or a course built to your curriculum.

The table sets the two side by side on the mechanics a buyer can check from each vendor's own pages. It is a comparison of stated design, not of learning outcomes, because we know of no independent study that compares the two.

Two rows deserve a closer look. The lesson-source row is the real trade-off. Skye's fixed bank means a school knows, in advance, every lesson a pupil can see on screen, even though Skye chooses its own spoken explanations and pacing within a session, and that is a safeguarding and quality strength. Dilr Academy's ability to build a course on request means it can teach a topic nobody has written yet, but the institution has to review that course rather than rely on a vendor-written bank. Our explainer on what Dilr Academy is walks through how a course is assembled and which courses are already built.

The teaching-method row is closer than it looks. Both tutors are designed to make the learner do the reasoning. Skye does it by asking a pupil to talk through each step of a scaffolded lesson, with hints and prompts along the way. Dilr Academy does it by showing the idea with a live diagram, asking before it tells, and moving on only once the learner has understood. Our guide to the Socratic AI tutor explains why that design matters and how to test for it, and our adaptive learning guide covers what knowledge tracing does and does not do.

The same diagnostic logic underpins our AI placement diagnostic, which starts from the problem a team is measured on and only then asks which tool fits it.

What does the evidence say about Third Space Learning?

The main independent evidence on Third Space Learning is an Education Endowment Foundation randomised trial of its earlier human-tutor model, run from September 2014 to May 2015, which found no evidence of impact on Key Stage 2 maths. We found no independent trial of Skye itself. The company's newer outcome figures are its own, and its impact page measures check-in against check-out questions within a session rather than attainment.

The trial is worth reading in full rather than in summary. The EEF Affordable Maths Tuition project page records a randomised design across 64 schools and 600 pupils. The pupils were Year 6, working at Key Stage 2 level 3 or an insecure level 4, and received weekly online tuition over 27 weeks from trained maths graduates in India and Sri Lanka. The key conclusion is direct: the evaluation found no evidence that the intervention had an impact on Key Stage 2 maths compared with business as usual teaching and support in Year 6.

Three caveats from the same page belong next to that result. First, it tested a human-tutor model in an early phase of the company's development, not Skye. Second, teachers were largely positive and reported that it appeared to improve pupils' comprehension, verbal fluency and confidence, outcomes the trial did not measure. Third, the process evaluation found that some pupils felt interrupted because the remote tutor could not see they were still working on a problem, and that the talking format worked better when pupils were spaced out in larger rooms. Those are design lessons any voice tutor, Skye or Dilr Academy, has to answer.

Third Space Learning's own figures are of a different kind. Its impact page reports a 2024/25 impact report covering 101,025 online one-to-one sessions, 9,364 pupils and 481 schools, and describes one of its measures: check-in questions at the start of a session and check-out questions at the end. It also asks pupils how their confidence changed. That shows whether a pupil was more secure on a concept at the end of a session. It does not show a change in SATs or GCSE results, and a buyer should not read it as if it did.

The broader evidence on tutoring is more encouraging. The EEF Teaching and Learning Toolkit on one to one tuition estimates approximately five additional months' progress on average, with short, regular sessions of about 30 minutes, three to five times a week, over up to ten weeks appearing to give the best results. The Toolkit defines tuition as support from a teacher, teaching assistant or other adult; it notes that studies involving digital technology show broadly similar effects, but it is not an estimate for AI tutors specifically. The Toolkit also sets the condition that should frame any pilot of either product: "Tuition is more likely to make an impact if it is additional to and explicitly linked with normal lessons."

The fair summary is that we found no independent attainment evidence for either AI tutor today, and Dilr Academy publishes no learner outcome statistics at all. That is not a reason to avoid both. It is a reason to run your own pilot with a comparison group, which we cover in the final section.

How do Skye and Dilr Academy handle safeguarding and pupil data?

Third Space Learning publishes a school-facing safeguarding and data model for Skye: every session is recorded and transcribed, concerns are flagged on a five-level scale with the most severe escalated to the school's designated safeguarding lead, and a due diligence pack includes a template DPIA. Dilr Academy's safety design covers sandboxed rendering and tenant isolation, but its live page does not describe an equivalent school safeguarding pack, and a school buyer should weigh that gap.

This is the section where Third Space Learning is clearly ahead for a school buyer, and we will not pretend otherwise. Its safeguarding page sets out a red-flag policy that sorts concerns from 1, the most severe, to 5, the least severe. A category 1 concern triggers review in near-live time before escalation to the school's designated safeguarding lead; categories 2 to 5 are reviewed at the end of the day and passed to the school where needed. Schools can request session recordings and transcripts at any point.

Its trust and compliance page goes further on data. Third Space Learning states that it acts as a data controller rather than a processor, because Skye exercises its own teaching judgement and the company holds its own safeguarding duties, so the school does not sign a data processing agreement. The school remains controller of the decision to use Skye and of which pupils take part. The same page says voice is transcribed in the moment so Skye can understand and reply, those transcriptions are not stored by the AI model used, recordings and session transcripts are kept on a rolling three-year schedule, and pupil data is not used to train future models. It also lists compliance claims against the DfE standards and the Children's Code. Those are the vendor's own statements, and a trust's DPO should test them against the documents rather than the badges.

Two documents frame the school's own review, whichever supplier it picks. The DfE's generative AI product safety standards, first published on 22 January 2025 and updated on 19 January 2026, are written for edtech developers and suppliers, and the DfE says schools and colleges may also find them helpful when judging which AI products are safe for use in education. And Keeping children safe in education is statutory guidance for schools and colleges, so the school keeps its own safeguarding responsibility. Third Space Learning states that it also holds safeguarding duties of its own, and its escalation process supports the designated safeguarding lead rather than replacing them.

Dilr Academy's position is narrower, and we state it as the live page does. Every interactive element is a validated template rendered in a sandbox, with tenant isolation, so there are no hallucinated interfaces. That protects against one class of risk, an AI inventing a screen. The live page does not describe a recorded-session policy, a red-flag scale or a pre-written DPIA. Dilr Academy is licensed to institutions for their own cohorts, and any school or college considering it for under-18 learners should run the full review in our safeguarding checklist before a pupil logs in.

The same review discipline sits inside our AI operating model work, which names an owner for every AI decision before a tool goes live.

Which buyers should stay with Third Space Learning?

Primary and secondary schools that need a defined maths intervention cohort, from Year 3 to GCSE, should usually stay with Third Space Learning. Skye is built for that exact job, its lessons are written by former teachers, its safeguarding model is published for schools, and its fixed annual pricing suits unlimited sessions. Switching to a general course-building tutor would trade those strengths for flexibility a maths intervention does not need.

We would give the same advice to a trust with these features:

• the target is KS2 SATs or GCSE maths, and the cohort is already identified;
• the school wants every word a pupil sees to come from a vendor-written lesson;
• the designated safeguarding lead wants recorded sessions and a red-flag feed as standard;
• staff time is short, and a set-up the school controls step by step, with progress reports, is the right level of control.

Third Space Learning's pricing model, which the company describes on its how it works page as unlimited tutoring for one fixed annual cost, also suits a school that wants to run as many sessions as a cohort needs without counting them. A buyer should take the current figures from the vendor directly.

If that profile is yours, the better use of your time is to make the intervention work: tie Skye sessions to the maths being taught that week, protect the session slots, and measure attainment against a comparison group. That is the EEF condition quoted above, applied to the product you already have. Our AI teacher buyer's guide covers how to set those measures before a contract renews.

What is the best Third Space Learning alternative in 2026?

The best Third Space Learning alternative in 2026 depends on the job. For school maths homework and practice, Sparx Maths, MathsWatch or Dr Frost Maths are closer like-for-like options. For free teacher tools and a large library, Khanmigo is strong, although its pupil tutor is not available to UK schools. For courses built on demand across subjects and languages, licensed to a college, university or training provider, Dilr Academy is the stronger fit.

We set the choice out as a short sequence rather than a ranking, because a ranking would pretend these tools do the same job.

Each option wins somewhere specific.

• Sparx Maths, MathsWatch and Dr Frost Maths are homework and practice platforms, as our Sparx Learning alternative guide sets out in detail. If the gap is consistent maths practice rather than one-to-one tuition, they are closer to the job.
• Century Tech describes personalised learning in English, maths and science across key stages, for a school that wants one personalised platform across subjects.
• Khanmigo offers free teacher tools and a large library, but our Khanmigo alternative guide explains why its pupil tutor is not available to UK schools today.
• Dilr Academy fits a college, university or training provider that needs a course nobody has written yet, in more than one language, taught Socratically with mastery tracking.

The government's own direction of travel is also worth watching. On 16 April 2026, DSIT and the DfE invited bids for AI tutoring tools for Years 9 to 10 in English, maths, science and modern foreign languages, with up to 8 companies forming a Pioneer Group and the aim of national availability from 2027. Any school buying a tutor now should expect the market, and the safety bar, to move during the contract.

For the broader map of where AI pays across education, see our education sector guide.

How should a school or college pilot an alternative to Skye?

A school or college should pilot an alternative to Skye on one defined cohort for about one term, with a comparison group, a named owner, and sessions tied to the lessons being taught that week. Measure attainment on a test you already trust, not the tutor's own in-session scores, and finish the safeguarding and data review before any learner starts. That gives a decision based on your learners rather than vendor claims.

A practical pilot has six parts.

1. Pick one cohort and one outcome. A Year 12 resit group, an adult numeracy class or a foundation-year module, measured on an assessment you already run.
2. Keep a comparison group. Without one, you cannot tell the tutor's effect from a good term.
3. Link sessions to lessons. Schedule short, regular sessions that follow the week's teaching, which is the EEF condition quoted earlier.
4. Finish the review first. Complete the DPIA, the safeguarding check and the data-sharing terms before a learner logs in.
5. Watch the room. The EEF trial found the talking format worked better when pupils were spaced out, so plan the space and the headsets.
6. Decide on attainment. Use your own test results, not session check-out scores, to decide whether to scale, switch or stop.

If you want help designing the pilot itself, our AI placement diagnostic starts with the outcome you are measured on, and our AI execution office can carry the delivery once a pilot is agreed. Our approach page explains how we keep a pilot tied to an owner and a measure.

Frequently asked questions

Does Dilr Academy replace Third Space Learning for a maths intervention cohort?

Dilr Academy does not replace Third Space Learning for a school maths intervention cohort. Skye is built for KS2 and GCSE maths intervention, with a fixed bank of teacher-written lessons and a published school safeguarding model. Dilr Academy builds courses on demand across subjects and languages for institutions. A school whose need is a defined maths intervention is usually better served by Skye.

The two can sit side by side in a trust with mixed needs, for example Skye for a GCSE maths cohort in its schools and Dilr Academy for a course the lesson bank does not cover.

Can parents buy Dilr Academy for their child?

Parents cannot buy Dilr Academy for their child today. Dilr Academy is licensed to institutions, which its page names as universities, colleges, schools and training providers, and it is not sold to individual learners. Third Space Learning also serves schools rather than families directly, so neither product is a home tutoring subscription. A parent looking for direct access should start with a consumer tutoring service instead.

Our AI teacher buyer's guide sets out how schools, universities and training teams buy differently.

Want to see a course built for your curriculum? Explore Dilr Academy, read the best AI education tool comparison, see our DATS methodology, or browse more AI strategy guides for education buyers.

Written by the Dilr.ai engineering team, practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed, or contact us about your own cohort.