The Mind

Key insights

The teacher is the paying customer; the student is the user; the parent is the one to reassure. MagicSchool has reached 5 million users because approximately 50% of US teachers have used it. Teachers pay $15 to $20 a month willingly because the return on investment is substantial: grading, feedback, and curriculum generation are handled by the platform. Students disengage after homework. This validates Outclas's B2B-to-institution positioning, and also indicates that Class Assistant is more strategically important commercially than I have been treating it. In school sales conversations, the presentation order should be teacher pain points first, student outcomes second.

Exam preparation is structurally a campaign, not a platform. A student preparing for TMUA or ESAT uses the product intensively for 6 weeks and then leaves. Sustained 4 to 5 days a week engagement (the metric a16z identifies as the one investors actually evaluate) is not achievable on a single high-stakes exam. The student is transient. Three durable responses: make the institution the durable relationship layer through annual renewable contracts; expand exam coverage so a single student remains a user across multiple years (TMUA at 18, A-levels at 17, GCSEs at 16); or position Parent Mentor as a continuous coaching subscription that endures beyond the exam window.

Outcome data is the moat, not the technology. Large language models are commoditised. Khan Academy could build a UK admissions tutor within a quarter. What competitors cannot replicate quickly is a multi-year, statistically meaningful track record of score improvements across hundreds of students. The 4 of 4 Oxbridge offers from the beta cohort is the foundation of that defensibility. The defensibility itself is built on the next 50, 100, then 500 students with measurable before-and-after results.

The Parent Mentor must be measured, not assumed. No mainstream competitor has established a substantive parent-facing layer. The one-brain-three-bodies positioning depends on this being real. If parents receive logins but never use them, the differentiation reduces to messaging without substance. Parent engagement must be instrumented from the first school deployment.

Multi-modal delivery is the wedge inside the exam-prep niche. Outclas currently sits in the "worksheet generator" category, which schools adopt but students tolerate. Voice-based worked examples (text-to-speech explaining solutions step by step) or short video walkthroughs would meaningfully differentiate the product without requiring celebrity deepfakes. Within the exam context, where students are already motivated, this represents a genuine engagement breakthrough.

UK exam preparation carries a structural distribution advantage that US edtech does not have. American AI edtech companies must partner with Pearson or McGraw-Hill for distribution. The UK does not operate this way. Exam boards (Cambridge Assessment, OCR, AQA) own the curriculum but do not gatekeep distribution. This advantage should appear in any fundraising materials.

Things to test

• Instrument the parent portal weekly-active rate at Brighton. Target: above 20% after 4 weeks. If the rate is lower, redesign Parent Mentor (probably toward a weekly digest email rather than a login portal) before the second school deploys.
• Build a voice-based worked-example feature for one ESAT or TMUA question type. Measure: does student return rate increase? Does time-on-task increase?
• Capture before-and-after TMUA and ESAT mock scores for every OxbridgeAITutor user in this cohort. Aggregate into a published outcome report by the end of the cohort.
• Run a £79 to £99 per month direct B2C subscription for OxbridgeAITutor in parallel with the institutional sales channel. Measure: signups, conversion, retention, and unit economics, compared against the same period of B2B effort.
• In the next school sales conversation, lead with the Class Assistant value proposition (teacher administrative pain) rather than AI Tutor (student outcomes). Observe whether the meeting progresses further.

Things to build into the plan

• Make A-level expansion explicit in the public plan after admissions-test preparation. (Reflected in the Plan section above as of 2026-06-10.)
• Treat Class Assistant as the commercial centrepiece in the sales narrative, not AI Tutor. AI Tutor is what students value most; Class Assistant is what teachers purchase.
• Add an outcomes page (separate from the homepage) where each cohort's measurable improvement is reported with proper statistical rigour. This is the artefact that converts an interested investor into a serious conversation.
• Introduce a B2C product line via OxbridgeAITutor direct subscription as a parallel revenue stream while B2B sales cycles take 6 to 12 months. Same product, different go-to-market.
• Prioritise outreach in this order: independent schools (Oxbridge-feeder) first, sixth-form colleges second, tutoring agencies third, state schools last.
• Reference Alpha School's published cohort outcomes (students performing at the 99th percentile nationally in the US, with approximately 2 hours of AI-directed instruction per day at $40,000 annual tuition) as a concrete external example of what AI-driven tutoring can achieve at the upper end. Useful in both sales narratives and investor materials to establish the ceiling for AI-tutor performance.

What I am watching for

• Whether Khanmigo, Synthesis, or a well-funded US player expands into UK admissions preparation. If they do, the window of unique positioning narrows considerably.
• Whether multi-modal AI tutors (voice, video, avatar-based) launch in exam preparation specifically. The first to execute this well in UK admissions secures a meaningful competitive advantage.
• Whether public secondary schools and FE colleges develop earmarked AI budgets at scale. If they do, the buyer pool expands beyond the independent and sixth-form niche.