Business Model and the Moat
You have spent many chapters learning how to build a tutor that actually teaches. This chapter asks a different, equally important question: how does this thing survive as a business? Two things decide that. First, your business model — who pays you, how, and how often. Second, your moat — the thing that stops a competitor (or the very company whose model you are renting) from copying you overnight. Let's define both in plain words and then connect them.
A moat is the old castle idea: a ditch of water around the castle that makes it hard to attack. In business it means any durable advantage that makes you hard to replace. Without a moat, the better your idea, the faster it gets copied.
28.1 The trap you must avoid: the "thin wrapper"
A large language model (LLM) is the kind of AI that powers chatbots — a system trained to predict and produce human-like text. Companies like OpenAI, Anthropic, and Google rent access to theirs through an API (Application Programming Interface — a doorway that lets your software send a question to their model and get an answer back).
A thin wrapper is a product that is little more than a clever instruction plus a pretty screen sitting on top of someone else's model. The problem: anyone can call the same doorway. Your only profit is the small gap between what the API costs you and what you charge — and that gap is vanishing. The price of running these models fell roughly 80% from 2023 to 2025. Worse, the model makers ship competing features straight to the public for free (OpenAI's "Study Mode", Google's NotebookLM).
28.2 The three ways to get paid
There are three classic shapes for an education-technology business. They are not just billing choices — each one demands a different product.
| Model | Who pays | Pace & feel | What it needs to build |
|---|---|---|---|
| B2C subscription (business-to-consumer) | The learner or parent, directly | Fast to sell, one person decides, but people quit easily (high churn) | Delightful self-serve app, parent progress view |
| B2B licensing (business-to-business) | A school, university, or company | Slow sale, a committee decides, but contracts last 1–3 years and renew | Teacher/admin dashboards, reporting, integrations |
| B2B2C | An institution buys it for its learners | Institution pays; learners use it free | Both: institutional contracts and consumer-grade engagement |
The money behaves differently too. A useful measure is LTV:CAC — Lifetime Value (total money one customer brings over their whole relationship) compared to Customer Acquisition Cost (what you spent to win them). Business-to-business education runs around 8–10× there, while direct-to-consumer is around 5–7×, because long institutional contracts are worth more per signup. Many winners blend models: Coursera leaned more on universities and employers as solo-learner numbers fell after the pandemic.
28.3 "Bring your own content" — teach from the learner's own material
Instead of only teaching your curriculum, you can let learners or institutions upload their own material — lecture slides, notes, a textbook — and have the tutor teach strictly from that. The technology is Retrieval-Augmented Generation (RAG): the AI answers from the uploaded sources rather than its fuzzy memory, which sharply cuts made-up answers and lets it show citations pointing to the exact page.
This is powerful for two reasons. It builds trust (a nervous student can verify every claim against the source), and it gives instant relevance — nobody waits for you to build content for their specific course. It also complements your own material instead of replacing it: your expert-verified curriculum is the backbone, and the upload feature handles the long tail of "the exact thing my professor assigned." Google's NotebookLM is the well-known example — upload sources, then chat with citations, quizzes, and flashcards grounded in them.
28.4 What actually makes a tutor defensible: four stacked moats
To move from a thin wrapper to a "thick" product, you stack advantages that grow stronger the more the product is used. That last part is the magic: a moat that compounds can't be copied by a newcomer because they're starting from zero today.
THIN WRAPPER THICK PRODUCT
+--------------+ +---------------------------+
| prompt + UI | --> | prompt + UI |
| on rented | | + learner-model moat |
| model | | + data/feedback moat |
+--------------+ | + content moat |
copyable in a day | + workflow lock-in |
+---------------------------+
gets stronger every session
- The learner-model moat (the strongest one). Over months you build a private map of exactly what each student knows, has forgotten, and should see next — assembled from knowledge tracing (continuously estimating mastery of each skill) and spaced repetition (resurfacing each item just before it fades). A rival starting fresh literally cannot reproduce a returning user's years of history. The model vendor has the foundation model; it does not have your student's long-term mastery record.
- The data/feedback moat. Every answer, mistake, and reaction to a hint quietly improves your scheduling and your content. Use makes the product better, which attracts more use.
- The content moat. Curated, expert-verified material mapped to the real exam blueprint or curriculum, with citations. Hard to copy at quality.
- Workflow lock-in (becoming the "system of record"). When the learning genuinely lives in your product — progress, schedules, cohorts, teacher dashboards, links into the school's other systems — leaving means throwing away years of history and retraining everyone. That switching pain is operational, not just a contract clause.
28.5 Niche-first beats "teach everything"
It is tempting to build a "learn anything" tutor to chase the biggest possible market. Resist it. A horizontal tutor has no opinion about how chemistry versus the bar exam should be taught, so it competes head-on with free general chatbots and wins on nothing. A niche-first product picks one domain and builds the whole experience around that domain's real workflow, vocabulary, and the way success is actually measured.
Choose a niche where (a) the outcome is measurable and high-stakes — a test score, a certification, fluency; (b) learners already pay today; and (c) generic chatbots do the job badly. Vertical, focused software keeps customers far better (91–96% retention) than broad horizontal tools (78–85%), and the proof shows in winners like ELSA Speak (pronunciation only) and Photomath (math photo-solving) — each the best at one job, not a worse ChatGPT.
28.6 Prove the outcome
In a high-stakes niche, the strongest marketing and the strongest retention driver are the same thing: measurable results. Khanmigo, Khan Academy's Socratic tutor, grew from 40,000 to 700,000 users across 380+ school districts in a single year — not by having a better raw model, but by embedding into school workflows and showing pilot gains of roughly 1.4 grade levels in math. Publish your score lift. Proof of outcome is what generic chatbots cannot casually claim.
- A clever prompt on a rented model is a "thin wrapper" — easily copied and squeezed as inference prices fall; the model makers will ship the generic version for free.
- Pick your business model (B2C, B2B, or B2B2C) from the unit economics, not gut feel — each one dictates a different product surface (parent vs. teacher vs. admin dashboards).
- Your four real moats are the learner model, the data/feedback loop, expert-verified content, and workflow lock-in — and the learner model is strongest because it compounds and can't be reproduced from scratch.
- "Bring your own content" with grounded, cited answers complements your curriculum, builds trust, and is a quiet wedge into institutions.
- Win one measurable, high-stakes niche and publish the outcomes — being the best at one job beats being a worse general chatbot.