Core Conclusions
AI education is not a single application but a composite value chain layering "content rights + learning data + learning workflows + model inference + distribution channels." The companies that can actually form a durable profit pool are not the "chat-style learning tools" that merely call a large model, but those that own authoritative content, assessment systems, school/enterprise entry points, and long-term learning-behavior data. As a result, publishing-and-assessment vendors, LMS/SIS, academic-integrity platforms, and enterprise learning platforms sit closer to the long-run winners than most pure AI front ends.
The scenarios that already generate real, verifiable revenue cluster into five categories: language-learning subscriptions (Duolingo, Speak, ELSA), enterprise AI-skills training and corporate learning platforms (Udemy, Docebo, Coursera for Business), testing and assessment (Pearson Study Prep, Duolingo English Test, Turnitin / online proctoring), AI add-on modules inside existing school software (PowerSchool, Instructure, Blackboard), and China's AI learning hardware / smart education (iFlytek learning machines, smart-education solutions).
What commercializes first, with the highest revenue certainty, is not the "universal AI tutor" but products that "embed into existing payment structures." For example, Duolingo packs AI features into subscription tiers; Pearson embeds AI questions, Study Prep, and teacher-assessment tools into its digital courses and assessment systems; Udemy/Docebo make AI part of enterprise seats and learning-platform add-on value; and Turnitin turns AI transparency and originality detection into a procurement line item for educational institutions.
The scenarios still stuck in pilots, concepts, subsidies, or "free-expansion" mode are mainly general-purpose campus copilots, district-wide K-12 AI tutors, and open-ended chat learning assistants for minors. ChatGPT Edu, Claude for Education, Google Gemini for Education, and Microsoft's education-edition Copilot are all entering campuses quickly, but most cases still emphasize deployment, pilots, training, discounted pricing, or giveaways rather than mature, stable, replicable, high-margin education products. Khanmigo on the district side also still carries an obvious nonprofit / partnership-expansion flavor.
The profit pool will not mainly stay with pure model companies in the near term. Model vendors and cloud providers will capture inference and seat revenue, but the high-repurchase, high-retention, high-switching-cost value of the education industry is more likely to stay with the companies that control the teaching process, the assessment process, and the procurement relationship—such as Pearson, PowerSchool, Instructure, Turnitin, Docebo, iFlytek, and some enterprise learning platforms.
AI's first-order impact on the education industry is not to immediately replace schools or teachers, but first to raise learning efficiency, compress content and tutoring costs, and broaden access to high-quality coaching. Anthropic's research on education scenarios shows teachers prefer "augmentation" over "automation"; OpenAI's collaboration with the AFT also centers on teacher training rather than teacher replacement.
In the near term, AI tutors look more like "low-cost, high-frequency, 24/7 supplementary tutors" than full substitutes for high-priced human tutors. The reason is that high-stakes scenarios still require trustworthy content, outcome verification, parental trust, minor safety, error correction, and contextual judgment; this is also why Khanmigo, Claude for Education, and OpenAI Study Mode all emphasize "Socratic guidance, teacher supervision, and step-by-step learning."
AI has already visibly disrupted traditional low-end question banks, homework answers, and homogeneous content platforms. Chegg's Q2 2025 revenue fell 36% year over year, subscription revenue fell 39%, and subscribers fell 40%, with management explicitly attributing the traffic decline to Google AI Overviews; this shows the "static answer bank" has been the first to be reconstructed by general AI and search AI.
AI-native challengers are accelerating into language learning, learning companionship, enterprise training, and skills assessment. Speak closed a $78 million Series C at a $1 billion valuation in late 2024; Preply raised $150 million at a $1.2 billion valuation in early 2026; ELSA keeps expanding enterprise and school partnerships; and Sana, Workera, and 360Learning are moving into high-ARPU customers on the enterprise-learning and skills-graph side.
The sub-sectors with the greatest revenue elasticity are language learning, enterprise AI reskilling, exam prep/grading, and AI markup modules inside existing LMS/SIS. Among them, language learning and enterprise training show the strongest willingness to pay; the school side, while carrying a high price per customer, has long procurement cycles, high compliance requirements, and high verification thresholds.
The sub-sectors with the best margins are not necessarily the "coolest" AI tutors but often the "AI add-on module + existing channel + low sales cost" model. Companies like Pearson, Turnitin, PowerSchool, and Instructure/Blackboard hold existing customer relationships and system entry points, so AI can lift ARPU; pure AI education apps often face higher customer-acquisition costs and more fragile retention.
On valuation, Duolingo and some large AI-platform companies already reflect "AI education" expectations fairly fully; Pearson, Docebo, and some Chinese education-tech and assessment-infrastructure companies may still hold expectation gaps. Conversely, although Chegg's valuation is extremely low, this looks more like a "structurally impaired discount" than a simple AI-beneficiary re-rating.
The most important catalysts over the next 12-24 months are not "another AI feature launch" but four quantifiable indicators: AI paying-user counts / add-on-seat penetration, learning-outcome verification, school/enterprise renewal rates, and large-scale procurement cases under education-regulation and data-security thresholds. The PowerSchool data-breach incident has already proven that the "safety and trust cost" of education AI directly affects where the profit pool lands.
Value Chain Landscape and Profit-Pool Attribution
From an investment standpoint, AI education can be divided into six layers: content/question banks, assessment/certification, model and cloud, learning products, education infrastructure software, and compliance and security. What truly determines long-term profitability is not just model capability, but who controls the teaching process, who owns trustworthy content, who holds the procurement relationship, and who accumulates learning data. Public information shows that even after Pearson embedded AI into Higher Education, assessment, and teacher tools, its profit still mainly comes from assessment and virtual schools; Turnitin, PowerSchool, and Instructure/Blackboard make AI an extension of their existing systems rather than starting from scratch.
Value-chain position Sub-segment Core products/services AI demand drivers Revenue model Main customers Content moat Data moat Channel moat Regulatory/security moat Margin profile Representative companies Listed/private Benefit intensity Investment elasticity Current view Upstream Educational content and rights Textbooks, question banks, exercises, course rights Needs trustworthy corpus, explainable answers, exam alignment Licensing, subscription, Inclusive Access Schools, students, enterprises High Medium Medium-high Medium-high Medium-high Pearson, McGraw/Cengage Pearson listed High Medium-high Platform-type core Upstream Question banks and exam blueprints Standardized question banks, essay scoring, speaking scoring, item authoring High-stakes exams need authoritative answers and scoring calibration Per-exam / per-school / per-question-bank licensing Exam bodies, schools, exam-prep firms Very high High High Very high High Pearson, Turnitin, Jiafa Education Listed/private High High Real-money scenario already verified Midstream Model and education-inference layer General LLM, education safety layer, RAG Generating explanations, dialogue, feedback, speech Seats, API, Edu editions Universities, enterprises, schools Low Medium Medium Medium-high High model margins but intense competition OpenAI, Anthropic, Google, Microsoft Private/listed Medium-high High Makes infrastructure money, not necessarily the whole industry's profit pool Midstream AI tutor 24/7 Q&A, Socratic tutor, learning companionship Lowers marginal tutoring cost, raises interaction frequency Subscription, family packs, school procurement Individuals, families, schools Medium Medium-high Medium High Good products can have high margins, but acquisition is heavy Duolingo, Khanmigo, Brainly, Quizlet Listed/private High Very high Easiest to break out on the consumer side, but moats diverge widely Midstream AI teacher assistant Lesson plans, homework, feedback, classroom activities Teacher time is scarce, prep cost is high Software add-on packs, seat fees Teachers, districts, schools Medium Low-medium High High Better than pure tools Google Classroom Gemini, Microsoft Copilot, Blackboard AIDA Listed/private Medium-high Medium-high Much of it still in free trial/bundling Midstream K-12 learning platform Adaptive learning, weakness diagnosis, learning paths Personalized learning and a data loop District procurement, per-student fees Schools, parents High Very high Very high Very high Medium-high iFlytek, PowerSchool, TAL Listed/privatized High High China and existing SIS/LMS entry points have the edge Midstream Higher-education learning platform LMS-embedded AI, course assistant, AI TA Universities must control compliance, academic integrity, course flow School license, module fees Universities, colleges Medium High Very high High Medium-high Instructure, Blackboard, OpenAI Edu, Claude for Education Privatized/private High Medium-high Existing vendors have the advantage over startups Midstream Exam prep Mock exams, diagnostics, adaptive drilling, essay/speaking scoring Students will pay for outcomes Subscription, per-exam, course packs Individuals, schools High High Medium High High Pearson Study Prep, Fenbi, New Oriental, Gaotu Listed High High Easiest to become a standalone paid product Midstream Language learning Speaking practice, video dialogue, pronunciation assessment High-frequency, global, strong willingness to pay Subscription, family packs, enterprise editions Individuals, enterprises Medium High Medium Medium High Duolingo, Speak, ELSA, Preply, Memrise Listed/private Very high Very high The most mature consumer-paid scenario in AI education Midstream Enterprise training and LXP/LMS AI skills training, role-based learning paths, course generation New enterprise AI budgets and reskilling needs Per-seat, ARR, implementation fees Enterprises, governments Medium High Very high Medium-high High SaaS margins Udemy, Coursera, Docebo, 360Learning, Sana Listed/private Very high Very high The enterprise side is the real big-B profit pool Midstream Skills assessment and skills graph AI skills testing, organizational skills maps Enterprises need "can they do it," not just "did they learn it" Enterprise contracts, assessment packs Enterprise HR/L&D Low-medium High Medium-high Medium High Workera, Docebo+365Talents Private/listed Medium-high High Could become an increment to enterprise learning budgets Downstream LMS/SIS/education infrastructure software Learning management, student information, school-home communication AI must embed into daily workflows to earn long-term payment Annual fees, module fees, implementation fees Schools, districts Low Very high Very high Very high High PowerSchool, Instructure, Blackboard, Clever Private/private/private Very high Medium-high One of the strongest channel moats Downstream Academic integrity and proctoring AI detection, originality, online proctoring Rising demand from AI cheating and remote exams Add-on, institutional license, per-exam Universities, K-12, certification Low-medium Medium-high High Very high High Turnitin, Honorlock, Meazure Private High Medium-high High demand, but high algorithm controversy and reputational risk Downstream Education data security Identity, permissions, privacy, and compliance Minor data and school-system security demands are high Security modules, implementation fees Schools, districts, universities Low High High Very high Medium-high PowerSchool, Clever, Microsoft/Google Private/listed Medium-high Medium Not the largest revenue pool, but it decides who can scale Customer layer Schools/districts Administration, teaching, compliance, school-home communication Security, integration, teacher training, procurement cycles Project-based + subscription K-12, universities - - Very high Very high Depends on implementation incumbent vendors - High Medium Slow decisions, but strong retention Customer layer Enterprise customers AI reskilling, sales enablement, compliance Tied directly to productivity Annual seats, ARR Enterprise HR/L&D - - High Medium High Udemy, Docebo, Coursera, Workera Listed/private Very high Very high The clearest new budget Customer layer Individual learners Language, exam prep, career skills, learning companionship High-frequency, low ticket; relies on retention and brand Monthly/annual subscription / single-exam Consumer users Medium Medium-high Medium Low-medium Extremely divergent Duolingo, Speak, Quizlet, Brainly Listed/private High Very high Easy to break out, but also easiest for general AI to undercut on price Profit-pool ranking judgment:
The first-tier profit pool is most likely to stay with companies that hold trustworthy content + assessment capability + procurement relationships, such as Pearson, Turnitin, PowerSchool, Instructure/Blackboard, and iFlytek. The second-tier profit pool stays with enterprise learning platforms, because enterprise AI-reskilling budgets are shifting from "experiment" to "institutionalized seats." The third-tier profit pool comes from AI-native consumer products, but only those with high-frequency interaction and a strong brand can survive the risk of general-AI substitution—Duolingo, Speak, and ELSA fall into this group. Model and cloud vendors will benefit significantly, but they look more like the "pick-and-shovel sellers of education AI" than necessarily the controllers of the final learning relationship.
Business Models, Value Capture, and Cost Structure
The pricing of AI education products has formed a clear tiering: the consumer side looks at retention, the B side looks at seats and renewals, and the school side looks at procurement and compliance. Language learning fits monthly/annual subscriptions best; enterprise training fits seats and ARR; school software fits site licensing, per-student fees, and modular markups; and exams/proctoring commonly use per-instance fees or institutional procurement. The model truly worth watching over the long term is "AI turning from an efficiency tool into a separately chargeable product," not merely internal cost reduction. Duolingo, Pearson, Turnitin, Udemy, and Docebo have already crossed that line; many teacher tools and school copilots have not.
Pricing-Model Comparison
Pricing model Typical scenario Pros Cons Better-suited companies Monthly/annual subscription Language learning, AI tutor, writing assistance High repurchase, high margin, supports feature iteration Needs strong retention and low CAC Duolingo, Speak, ELSA, Memrise Family/multi-user packs Family learning assistant, language learning Raises ARPU, lowers churn Account sharing may dilute per-user value Duolingo Per-student / per-school site K-12 platforms, LMS, school AI Predictable scale, long contract terms Long sales cycle, needs compliance and implementation PowerSchool, Instructure, Blackboard Per-teacher-tool fee Lesson-plan generation, grading, classroom-activity design ROI easy to prove, saves teacher time School budgets are sensitive, easily made free by suites Google Classroom, Microsoft Copilot, Blackboard AIDA Per enterprise seat / ARR Enterprise training, AI reskilling, LXP Large budget, strong renewals, expandable modules Needs continuous content updates and customer success Udemy, Docebo, Coursera, 360Learning Per course/certificate fee MOOCs, professional certification, exam training High ticket, outcome-oriented Completion rates and acquisition costs matter Coursera, New Oriental, Fenbi Per-exam / per-instance fee DET, essay scoring, speaking scoring, proctoring Tied to outcomes, stronger willingness to pay High compliance, high reputational risk Duolingo English Test, Pearson, Honorlock, Meazure Per AI call / usage-based fee API-type education tools, private deployment Strongly tied to usage Unfavorable for stable institutional budgets OpenAI, Anthropic, cloud vendors Value Capture and Budget Flows
Based on public pricing, procurement structures, and company disclosures, schools, universities, and enterprises typically allocate their AI-education budgets along the following logic: Schools/universities place more weight on learning platforms, assessment, data security, teacher training, and implementation services; Enterprises place more weight on content generation, skills graphs, role-based learning paths, seat management, and ROI analysis. Google's education-AI pricing shows the suggested prices for Gemini Education / Premium are $20 and $30 per user per month respectively; Microsoft's education-edition Microsoft 365 Copilot is priced at $18 per user per month; this shows the "AI seat fee" is becoming a standalone line item in school and university budgets.
Budget item School/university budget weight Enterprise budget weight Value judgment Content/question banks/rights 15%-30% 10%-20% High moat, high reuse Learning platform/LMS/SIS 20%-30% 15%-25% Main process entry, strong long-term stickiness Model calls/AI seats 10%-20% 10%-20% Cost gradually falling, but can be bundled by platforms Data security/privacy/audit 5%-15% 5%-10% Prerequisite for whether education AI can scale Teacher/admin training 5%-10% 5%-10% Directly affects adoption rate Assessment/integrity/proctoring 5%-15% Low A must-have in high-stakes scenarios Implementation services/integration 10%-20% 5%-15% Clearly higher on the school side than the enterprise side Hardware 0%-20% Low More important in China; mostly non-core overseas Where AI Most Easily Cuts Costs and Lifts Revenue
The places where AI most easily lowers cost are content production, question generation, first-pass homework grading, teacher prep, customer-service Q&A, enterprise internal course production, and course localization. Chegg explicitly disclosed that, by relying on AI usage, the company expects to cut more than $50 million of content and software development capex in 2026 versus 2024; Udemy also significantly reduced R&D and marketing-efficiency drag in 2025 and expanded its Adjusted EBITDA margin.
The places where AI most easily creates new revenue are premium language-learning subscriptions, enterprise AI-skills training, exam-scoring and proctoring add-on packs, and AI markup modules inside LMS/SIS. It is precisely these scenarios where customers will pay separately for "learning faster / taking exams more securely / less manual work / better management."
Three-Scenario Forecast
Dimension Conservative Base Aggressive Key assumption Schools cautious, regulation tightens, teacher burden rises Enterprise training expands, schools adopt embedded AI AI-tutor outcomes improve, platforms deeply integrate Student AI adoption High usage, low payment High Very high School AI procurement rate Low to medium Medium Medium-high Enterprise AI-training demand Medium High Very high AI-product payment rate Low Medium Medium-high Learning-outcome verification Limited evidence Partial evidence strengthens Clear improvement Customer-acquisition cost High Medium Medium-low Software revenue growth Medium-low Medium-high High Beneficiary segments Enterprise training, integrity tools, language learning Language learning, exam prep, LMS/SIS markups, enterprise LXP AI tutor, learning OS, enterprise skills graph Beneficiary companies Udemy, Docebo, Turnitin, Duolingo Pearson, Duolingo, Docebo, PowerSchool, iFlytek Duolingo, Speak, Pearson, Instructure, Workera Disrupted companies Chegg, low-end question banks, low-end content platforms Chegg, some homework-answer platforms, low-end exam prep Traditional tutoring, static question banks, some general-course platforms Main risks Minor privacy, outcome disputes, procurement delays Big-tech suite free-ification, weak renewals Security incidents, model errors, regulatory reversal Reshaping of Teacher, Tutor, Trainer, and Content Roles
Teachers, tutors, trainers, TAs, and instructional designers will not be replaced wholesale within the foreseeable horizon, but the structure of their work will change markedly: First, repetitive work gets automated; Second, the importance of supervision, error correction, evaluation, and context-specific design rises; Third, "human-machine collaboration ability" becomes a new labor skill in education services. Anthropic's study of 74,000 educator conversations shows teachers most often use AI for curriculum development, research, and student assessment; OpenAI's collaboration with the AFT on a national-level training program for 400,000 teachers also proves the industry logic is "teacher augmentation," not "teacher exit."
Deep Dive into Sub-Sectors
The table below compresses the 30 sub-sectors requested by the user into a single investment-judgment table. Scores are on a 10-point scale; higher scores indicate stronger commercialization certainty, moats, and investment trackability.
Sub-sector TAM direction How revenue forms Current commercialization stage Pricing model Source of moat Learning outcome/regulation 12-24m catalysts Main risks Attractiveness AI tutor Large Subscription/school procurement Commercialized but limited outcome verification Monthly subscription/school pack Dialogue data, brand, content safety Outcome evidence diverges, high minor-safety requirements Study Mode, Khanmigo, Claude learning-mode expansion Homogenized by general AI 8.0 K-12 AI learning platform Very large District procurement/per-student fees Early commercial, slow procurement Annual fee/school pack School data, channels, compliance Strictest regulation, long procurement cycle District-level cases, national/local digital-education policy Privacy and security incidents 8.2 Higher-education AI Large Campus license/LMS add-on module Rapid-deployment phase Campus site LMS, course flow, academic integrity Relatively high university acceptance OpenAI/Anthropic/Canvas expansion Academic-integrity disputes 7.8 AI teacher assistant Large Teacher seats/platform markup Many launches, payment still ramping Seats/bundling Workflow embedding Clear time-saving ROI, more indirect learning outcome Google/Microsoft/Blackboard feature expansion Free-ification price pressure 7.5 AI homework grading Medium-large School module fee Mature scenarios exist School module fee Labeled data, rubric High-stakes needs human review Deep LMS integration Misjudgment and fairness 7.2 AI essay and writing assistance Large Subscription/institutional license Already chargeable Subscription/Add-on Writing corpus, feedback loop Affected by both study-aid and cheating disputes Turnitin Clarity, writing-transparency tools Over-reliance on AI 7.4 AI math learning Very large Subscription/device/school procurement Strong commercialization Subscription/hardware Wrong-answer data, step-by-step reasoning Outcomes easier to quantify China learning machines / international math-tutor evolution Wrong-solution risk 8.1 AI science learning Medium-large Subscription/course packs Early to mid Course packs/subscription Experiment content, course resources Needs stronger knowledge accuracy Multimodal experiments/simulation Hallucination and experiment safety 6.8 AI programming education Large Subscription/course/enterprise training Faster commercialization Subscription/course fee Code assessment, project library Strong employment orientation Enterprise AI-dev training expansion Direct substitution by general AI 7.6 AI language learning Very large Subscription/B2B Most mature Monthly/annual subscription High-frequency interaction, brand, voice data Perceptible outcomes, relatively light regulation Duolingo/Speak/ELSA penetration Apple/Google translation substitution 9.0 AI speaking practice Large Premium subscription/enterprise English One of the most mature Premium subscription/enterprise edition Speech-assessment data Users pay to "dare to speak" Video dialogue and multilingual expansion Cost and retention 8.8 AI exam prep Very large Subscription/question bank/mock exam/grading Verified Subscription/single-exam pack Question banks and scoring rules Most outcome-focused, also strong repurchase Standardized exams and professional certificates upcycle Free-ification of general-AI answering 8.7 AI professional certification Large Course packs/question-bank subscription Verified Cohort classes + subscription Question banks, faculty, pass-rate brand Regulation and reputation matter China civil-service/accounting/law exams going AI Insufficient user trust in "pure AI" 8.1 AI enterprise training Very large Seats/ARR/implementation Verified Annual seats Customer relationships, content management, skills graph Measurable ROI Expanding enterprise AI budgets Macro budget volatility 9.0 AI skills assessment Large Enterprise contracts/assessment packs Mid-early Assessment packs/platform fee Skills graph and assessment data High value to enterprise decisions Rising AI-role-assessment demand Certification too slow 7.9 AI learning management system Very large Platform markup/module fee Verified ARR/module fee Workflow and data-migration cost Strongly relevant to both schools and enterprises Canvas/OpenAI, PowerBuddy Big-tech suite squeeze 8.8 AI student information system Large SIS upgrade/module fee Verified but cautious Annual fee/implementation Core master data Extremely high privacy requirements School-home Q&A, risk alerts Data breach 8.3 AI educational publishing Large Content licensing/digital subscription Verified Licensing/subscription Authoritative content and exam alignment Strong credibility Publisher-cloud-vendor partnerships Price pressure from general content generation 8.5 AI question bank and practice system Large Subscription/institutional license Verified Subscription/question-bank pack Historical question banks and behavior data Highly outcome-oriented Adaptive drilling and wrong-answer-notebook enhancement Commoditization 8.4 AI course generation Medium-large B2B tools/teacher seats Early Seats/tool fee Workflow integration Outcomes depend on teacher review Google/Blackboard/Docebo deepening Severe free-ification 6.8 AI education video and content production Medium-large Tool fee/platform bundling Early Seats/API Templates and distribution Indirect contribution to learning outcomes Faster multilingual localization Fast homogenization 6.5 AI academic integrity Large Institutional license/Add-on Verified Institutional annual fee Databases, institutional relationships High demand but big controversy AI-native writing-transparency products False positives and reputational risk 8.0 AI online proctoring Medium-large Per-exam/institutional procurement Verified Per-instance/contract Process, identity recognition, question-leak monitoring Extremely high compliance More remote exams Technical failures/litigation 6.9 AI study-abroad and admissions advising Medium-large Consulting packs/subscription Mid-early Service packs Case library, application experience High value but high trust requirements Essay/school-selection/scholarship assistants Regulation and accuracy 6.7 AI career planning and interview coaching Large Subscription/enterprise edition Ramping fast Subscription/enterprise pack Scenario data, simulated feedback Outcome verification needs strengthening Expansion of AI recruiting and talent matching Substitution by general AI 7.3 AI education data platform Large Platform fee/implementation Mid-stage Platform fee Cross-system data integration Privacy and sovereignty critical District/university data-platform upgrades Complex procurement 7.7 AI education security and privacy Large Security modules/consulting Must-have, strengthening Project-based + renewal Compliance capability Strongest regulatory rigidity Upgrades after data-breach incidents Budget pressure 8.0 AI learning companionship Large Subscription/companion packs Growing fast Monthly subscription Emotional design and sustained interaction High risk for minors Multimodal companionship and characterization Safety and dependency issues 7.0 AI education hardware Medium-large Device sales + subscription More mature in China Hardware + service Channels, supply chain, device data Parents willing to pay, but ebbs fast too Learning-machine upgrade cycles Hardware inventory and price wars 7.5 AI education consulting and implementation services Medium-large Project-based/training Genuinely exists Project fee Delivery capability and school-enterprise relationships Not the core of the platform profit pool University/enterprise deployment wave Non-standard, labor-intensive 6.4 Sub-sector conclusion: The five sub-sectors most worth watching are: AI language learning, AI enterprise training, AI exam prep, AI learning management system, and AI educational publishing/assessment. Because they either already have real payment data, or hold content/data/channel moats, or can turn AI from "feature enhancement" into "a product customers will pay separately for." By contrast, purely conceptual AI tutors, free teacher copilots, and homogeneous content-generation tools, while strong on narrative, are weaker on both commercialization and moats.
Master Investment Table and Company Tiers
Master Investment Table
The table below puts listed, private, platform-type, pick-and-shovel, challenger, and disrupted companies in the same framework. Scores are subjective research scores: benefit certainty / earnings elasticity / regulatory risk / valuation attractiveness, out of 10.
Company Region/status Segment Core AI-education products/services AI benefit path or disruption path Key operating signals Category Score summary Duolingo US/listed Language learning Max, Video Call with Lily, AI speaking Direct beneficiary; AI enters subscription tiers, creating higher ARPU and stronger retention 2025 revenue $1.038 billion, 12.2 million paid users, Q4 DAU 52.7 million; company calls Max one of the most successful consumer AI products A Certainty 9 / elasticity 9 / regulation 3 / valuation 4 Pearson UK/listed Publishing + assessment + higher ed Study Prep, AI study tools, teacher-assessment tools Direct beneficiary + platform-type; uses authoritative content and assessment systems to mark up 2025 operating profit up 6%, Study Prep continues monetizing, AI learning tools improve learning outcomes A 8 / 8 / 4 / 7 Udemy US/listed Enterprise training/online learning AI Packages, AI Growth/Readiness Collections Direct beneficiary; enterprise AI reskilling brings ARR 2025 revenue $790 million, enterprise ARR $540 million, AI content learning time 700 million minutes A 8 / 8 / 3 / 7 Coursera US/listed Online learning/enterprise training AI courses, professional certificates, enterprise customers Indirect beneficiary; strong AI content demand, but pricing and retention still need verification Agreed in late 2025 to acquire Udemy, combined entity valued at $2.5 billion, betting on enterprise AI training B 6 / 8 / 3 / 7 Docebo Canada/listed Enterprise LMS/LXP AI-first learning platform, skills intelligence Direct beneficiary + platform-type Mid-2025 ARR $233 million, 2025 subscription-growth guidance ~11%, acquired 365Talents A 8 / 7 / 2 / 7 Microsoft US/listed Pick-and-shovel/campus workflow Copilot in Education Pick-and-shovel + platform-type; embeds AI into Office and school IT systems Education-edition Copilot academic pricing $18/user/month B 7 / 7 / 3 / 4 Alphabet / Google US/listed Pick-and-shovel/classroom tools Gemini for Education, NotebookLM, Classroom Gemini Pick-and-shovel + platform-type Gemini Education $20/user/month, Premium $30/user/month; opened more AI tools to all education editions in 2025 B 7 / 7 / 3 / 4 OpenAI education business US/private Model + campus deployment ChatGPT Edu, Study Mode Pick-and-shovel + potential platformization CSU-system deployment covers about 460,000 students and 63,000 faculty/staff; Study Mode targets education scenarios A-private 7 / 8 / 4 / NA Anthropic education business US/private Model + higher ed Claude for Education Pick-and-shovel + potential platformization Launched April 2025; Northeastern campus-wide deployment; educator use cases center on curriculum development and assessment B-private 6 / 8 / 4 / NA PowerSchool US/privatized SIS/LMS/K-12 software PowerBuddy Platform-type beneficiary; very strong SIS data and school-home entry point Privatized by Bain in 2024 for $5.6 billion; serves over 60 million students; but data breach exposed risk A-privatized 8 / 7 / 8 / NA Instructure US/privatized LMS IgniteAI, embedding with OpenAI into Canvas Platform-type beneficiary Partnered with OpenAI to embed AI directly into Canvas, emphasizing privacy-first and interoperability A-privatized 8 / 7 / 4 / NA Blackboard / Anthology US/privatized restructuring Higher-ed LMS AI Design Assistant, Virtual Assistant Indirect beneficiary; large customer base, but heavy financial pressure Offered existing customers free trials through June 2026; corporate restructuring shows financial strain C 5 / 5 / 5 / NA Turnitin US/private Academic integrity/writing Originality, Clarity, AI detection Platform-type beneficiary Serves 16,000+ institutions, Clarity starting to enter secondary schools; clear AI-detection paid-add-on logic A-private 8 / 7 / 6 / NA Honorlock US/private Online proctoring AI proctoring, AI blocking, phone detection Direct beneficiary, but highly controversial Product emphasizes AI + human proctoring B-private 6 / 6 / 7 / NA Meazure Learning / ProctorU US/private Online proctoring Remote proctoring platform Direct beneficiary but high risk Runs "millions" of exams; California bar-exam incident caused litigation D-private 4 / 5 / 9 / NA iFlytek China A-share/listed Smart education + learning machines Learning machines, essay grading, speaking practice, smart education Direct beneficiary + platform-type 2025 smart-education revenue 8.967 billion yuan, up 24.04%; learning machines ranked No. 1 in high-end sales for five consecutive years A 9 / 8 / 5 / 7 New Oriental China US/HK/listed Exam prep + study abroad + adult education New education business, AI tools Indirect beneficiary; AI is mostly efficiency and product enhancement FY2025 revenue $4.90 billion, up 13.6%, with growth from the new education business; AI direct revenue not separately disclosed B 6 / 6 / 4 / 6 TAL Education China US/listed Learning services + learning devices AI-driven learning devices Direct-benefit potential; hardware + learning data The 2024 annual report already clearly placed AI-driven learning devices in the core business, but the 2025 AI financial contribution is not clearly itemized B 6 / 7 / 5 / 6 NetEase Youdao China US/listed Learning services + dictionary/hardware AI subscriptions, AI learning tools Direct-benefit potential; but limited disclosure granularity Public IR pages confirm ongoing disclosure, but current AI revenue line items need further verification C 5 / 7 / 5 / 6 Gaotu China US/listed Courses + exam prep AI courses, learning assistance Direct-benefit potential; but standalone AI revenue not clearly disclosed The company's 2025 annual report is disclosed, but AI contribution needs further breakout C 5 / 6 / 5 / 6 Fenbi China HK/listed Civil-service/professional exams Question banks, mock exams, AI interview/grading potential Direct-benefit potential; strong question-bank and civil-service data moat Current AI financial-contribution disclosure is limited and needs ongoing verification B 6 / 7 / 4 / 6 Offcn Education China A-share/listed Civil-service/employment training AI employment learning machine Narrative stronger than current verification Launched an AI employment learning machine in 2025, but the core business recovery is still fragile D 4 / 6 / 5 / 6 Jiafa Education China A-share/listed Smart testing AI item-authoring assistant, English listening/speaking computerized exams, inspection Platform-type beneficiary "AI+" covers exam-teaching-management-assessment-learning-research; 2025 revenue roughly flat, net profit declined B 6 / 7 / 5 / 7 Benesse Japan/listed K-12/family education AI learning, textbooks and tutoring Worth tracking; but limited public AI-monetization data AI standalone revenue needs further verification C 5 / 5 / 4 / 6 Quizlet US/private Learning tools Q-Chat, AI study guides AI-native challenger One of the first AI tutors built on the OpenAI API; but financials undisclosed B-private 6 / 8 / 3 / NA Brainly US/Poland/private Homework help/AI learning companionship AI Learning Companion, Test Prep AI-native challenger Models around students, emphasizing COPPA/GDPR/CPRA B-private 6 / 8 / 5 / NA Khan Academy / Khanmigo US/private/nonprofit AI tutor Khanmigo AI-native challenger, but weak profit objective District partnerships, Enterprise Starter, free in some state programs B-private 6 / 7 / 4 / NA Speak US/Korea/private AI speaking practice AI Tutor AI-native challenger, one of the clearest direct-payment cases Raised $78 million in 2024 at a $1 billion valuation A-private 8 / 9 / 3 / NA ELSA Speak US/Vietnam/private Pronunciation assessment/enterprise English Pronunciation AI, Dashboard AI-native challenger Partners with Pearson, Oxford, schools, and enterprises B-private 7 / 8 / 3 / NA Preply Europe/private Human + AI language learning AI-enhanced human-tutor platform Human-machine collaboration challenger Raised $150 million in 2026 at a $1.2 billion valuation, 100,000+ tutors globally A-private 8 / 8 / 4 / NA Sana Europe/private, later acquired Enterprise learning/knowledge platform AI-native learning platform Enterprise-learning challenger Workday's planned acquisition shows its strategic value is recognized by mainstream HR software B-private 7 / 8 / 3 / NA Workera US/private Skills assessment Skills Intelligence Platform-type challenger Cases such as Siemens Energy, emphasizing verifiable skills data B-private 7 / 8 / 3 / NA Company Tiers and Investment Priority
Tier A: Core direct beneficiaries of AI education Duolingo, Pearson, Udemy, Docebo, iFlytek, Turnitin, PowerSchool, Instructure, Speak, Preply. These companies either have already turned AI into standalone paid products / add-on packs / ARR, or control key entry points to school and enterprise learning.
Tier B: Clear beneficiaries, but with valuation, regulatory, or competitive risk Coursera, Google, Microsoft, TAL, Fenbi, Khanmigo, ELSA, Workera, Jiafa Education. Their benefit logic holds, but they are either pick-and-shovel sellers, or their LMS/content markup is still being verified, or their disclosure is insufficient.
Tier C: AI mainly used for efficiency, weak near-term financial elasticity New Oriental, Benesse, Youdao, Gaotu, Blackboard/Anthology. Such companies are currently more about boosting efficiency, retaining customers, and maintaining existing competitiveness than adding a high-growth curve through AI.
Tier D: Strong AI-education narrative, but insufficient evidence of actual benefit Offcn Education, some teacher tools and free copilot products, some AI homework assistants. A common trait is many launches, many trials, strong narrative—but insufficient paying users, renewal rates, and learning-outcome verification.
Tier E: Companies potentially disrupted by AI-native tools or general AI Chegg, low-end homework-answer platforms, low-differentiation question banks, low-end content production, and some traditional online-course platforms; the 2U / some OPM models have also been proven fragile by the industry.
Scoring Model
Positive scoring model—suggested weights
Dimension Weight Direct AI-education revenue exposure 20% Content, data, and channel moats 20% User retention and payment conversion 15% Learning-outcome and security-compliance verification 15% Financial quality and margins 10% Market size and growth elasticity 10% Valuation reasonableness 10% Reverse risk-scoring model—suggested weights
Risk dimension Weight Insufficient user retention 20% Insufficient learning-outcome verification 20% Minor-privacy and regulatory risk 20% General-AI substitution risk 15% Excessive customer-acquisition cost 15% Overvaluation 10% Composite ranking suggestion
Rank Company Positive total Main reason 1 Pearson 83 Content + assessment + school/higher-ed channels in one, AI already starting to monetize 2 Duolingo 82 Strongest consumer AI payment and brand, but high valuation expectations 3 Docebo 80 Enterprise-learning platformization benefit, AI-first + skills intelligence 4 iFlytek 79 China education hardware + smart education dual engine, revenue already verified 5 Turnitin 78 Academic integrity is a must-have in the AI era, strong institutional channel and database moat 6 Udemy 77 Clear enterprise AI-training and ARR path, more upside imagination after the merger 7 PowerSchool 76 Scarce school master-data entry point, but high security risk 8 Instructure 75 Canvas is a key higher-ed workflow, high value from AI embedding 9 Speak 74 Extremely strong AI-speaking-practice productization, still needs verification of long-term retention and cost 10 Preply 73 Human + AI collaboration model is relatively resilient to general AI, but slightly lower margin as it is not pure software In-Depth Analysis of Key Listed Companies
Duolingo (NASDAQ: DUOL) Duolingo is currently the clearest example of an education company moving along the logic of "AI feature to standalone paid product, then to core subscription." In 2025 the company had revenue of $1.038 billion, total bookings of $1.158 billion, and net income of $414 million; in Q4 paid subscribers reached 12.2 million and DAU reached 52.7 million. Management explicitly stated that Duolingo Max and its core feature Video Call with Lily are "one of the most successful consumer AI products," and plans to extend the core voice-conversation capability to larger subscription tiers in exchange for higher DAU and broader penetration. The moat comes from brand, gamified retention, voice/behavior data, and global distribution; the risk is that the valuation already reflects AI expectations significantly, and the company is proactively lowering near-term monetization friction, so 2026 margins may come under pressure. Research conclusion: strong beneficiary / high certainty / valuation on the high side but still worth ongoing tracking.
Pearson (LSE: PSON) Pearson is the most typical case of a "traditional education content and assessment company being re-platformed by AI." The company's 2025 operating profit grew 6%, and it continues to monetize Study Prep and AI study tools in higher education; official disclosures link its AI-powered study tools to improved learning outcomes, and Study Prep keeps expanding into international markets. More importantly, Reuters reported that about 80% of Pearson's 2025 operating profit came from assessment and virtual schools, meaning its profit pool does not depend on the low-moat content most easily replaced by general AI, but on high-trust, high-compliance assessment and school relationships. Research conclusion: strong beneficiary / platform-type winner / expectation gap exists.
Udemy (NASDAQ: UDMY) Udemy's key change is not "many AI features" but that its business structure is starting to converge toward subscriptions and enterprise customers. In 2025 revenue was $789.8 million, of which enterprise revenue was $524.1 million and ARR was $540 million; total gross margin was 66%, Adjusted EBITDA margin was 12%, consumer subscription revenue grew 44% year over year, and paid consumer subscribers reached 343,000. The company also launched AI Growth / AI Readiness packages, with over 700 million minutes of AI content learning in 2025, reflecting that enterprise AI-reskilling demand is real. Research conclusion: strong beneficiary / relatively large earnings elasticity / worth re-evaluating after the merger with Coursera.
Coursera (NYSE: COUR) Coursera's benefit path lies in "increased AI course supply + enterprise-customer growth + professional-certificate platformization." But the difference from Duolingo and Pearson is that AI is currently more about driving content demand and enterprise learning budgets than forming a high-stickiness proprietary teaching workflow. In late 2025 Coursera agreed to acquire Udemy in an all-stock deal, with the combined entity valued at about $2.5 billion, clearly signaling that management believes the industry needs scale, enterprise customers, and more subscription-type revenue to hedge single-platform pressure. Business Insider reported that Coursera's CEO said enrollment speed in AI-themed courses has risen from 1 person per minute in 2023 to 14 per minute in 2025. Research conclusion: medium beneficiary / high elasticity / needs continued verification of merger integration and retention.
Chegg (NYSE: CHGG) Chegg is the most typical negative sample of AI disrupting the education industry. In Q2 2025 the company had revenue of $105.1 million, down 36% year over year; subscription-service revenue fell 39%, and subscribers dropped to 2.6 million, down 40% year over year; management explicitly attributed the traffic decline to Google AI Overviews. Although the company is using AI to improve efficiency and expects to cut more than $50 million of content and software development capex in 2026 versus 2024, this is more of a "self-rescue cost reduction" that cannot mask the structural fact that the "static answer-bank model has been weakened by search AI and general AI." Research conclusion: likely disrupted / low valuation does not equal low risk.
Docebo (TSX/NASDAQ: DCBO) Docebo is one of the most trackable "AI-first beneficiaries" among enterprise learning platforms. The company disclosed that as of mid-2025 ARR reached $233.1 million, with 2025 subscription-revenue growth guidance of about 10.75%-11.75%; in 2026 it further strengthened skills intelligence and enterprise knowledge infrastructure by acquiring 365Talents and Zive. The key for such companies is that enterprise customers were already paying for LMS/LXP, and AI is not an extra education narrative but a direct lift to seat value, content-generation efficiency, and skills-matching capability. Research conclusion: strong beneficiary / valuation relatively debatable / clear platform logic.
Microsoft (NASDAQ: MSFT) Microsoft's role in education AI looks more like "pick-and-shovel seller + workflow platform." Its advantage lies not in standalone education content but in the embedded edge of Word, PowerPoint, Teams, OneDrive, campus IT systems, and identity management. The company has launched an academic-edition Microsoft 365 Copilot for education customers priced at $18 per user per month, effectively turning the AI seat fee into part of school IT budgets. For investment research, this line is real, but education is only part of its enormous platform, and AI-education revenue can hardly move the overall valuation on its own. Research conclusion: clear beneficiary / but education AI has limited impact on overall financial elasticity.
Alphabet / Google (NASDAQ: GOOGL) Google's education-AI path resembles Microsoft's, but it is closer to teaching scenarios in Classroom, Chromebook, NotebookLM, and the Gemini tab. Google Workspace for Education pricing shows Gemini Education at $20 per user per month and Premium at $30 per user per month, indicating that the school-side AI seat fee has begun to form an independent budget line. The issue is that Google is also releasing education-AI capabilities for free or at a lower threshold, which accelerates adoption but also compresses the space for standalone education apps. Research conclusion: platform-type beneficiary / education AI looks more like a defensive and ecosystem benefit.
iFlytek (SZSE: 002230) iFlytek is one of the few companies in the China market to turn AI education into large-scale revenue. The company's 2025 smart-education business revenue was 8.967 billion yuan, up 24.04% year over year; its learning machines ranked No. 1 in high-end learning-machine sales volume and sales value for five consecutive years, with features such as AI precision learning, AI essay grading, and AI speaking practice. Unlike overseas pure software, iFlytek combines hardware, content, local education channels, and government-education relationships, allowing it to sell AI education as a "device + software + school solution." Research conclusion: core direct beneficiary of China AI education / dual-engine drive of platform and hardware.
New Oriental (NYSE: EDU / HKEX: 9901) New Oriental's main logic is still the recovery of exam prep, study abroad, and the new education business, rather than AI driving on its own. The company's FY2025 revenue was $4.90 billion, up 13.6% year over year; the first half of FY2026 continued to grow, with management attributing the growth mainly to the "new education business" rather than explicit AI monetization. Its advantages lie in brand, channels, faculty, and study-abroad/exam-prep scenarios, but AI is more a tool to raise capacity and improve product form. Research conclusion: medium beneficiary / AI efficiency gains stronger than revenue elasticity.
TAL Education (NYSE: TAL) TAL's AI benefit path is mainly in learning services and AI-driven learning devices. The annual report has placed AI-driven learning devices into its core-business statement, showing it intends to pursue a "hardware entry + learning data + service subscription" route over the long run. The issue is that current public disclosure is still insufficient to break out AI revenue separately, so it is better placed on the "has potential but needs continued verification" list. Research conclusion: medium beneficiary / needs ongoing tracking of device sales, user retention, and service revenue.
NetEase Youdao (NYSE: DAO) Youdao has natural scenarios combining dictionary, translation, learning hardware, content, and AI technology, making it one of the few companies in China that could turn "AI-enabling learning tools" into real payment. But given the current availability of public materials, its AI-subscription revenue, AI-hardware margin, and user-retention breakdown are not sufficiently clear. Research conclusion: benefit logic holds, but financial verification is insufficient; need to focus on verifying standalone AI payment and hardware-software synergy.
Gaotu (NYSE: GOTU) Gaotu belongs to the group of "not-weak AI narrative, but AI financial contribution still needs verification." It naturally has live-streamed classes, exam prep, and adult education scenarios that let AI enter Q&A, homework grading, course generation, and learning planning, but current public disclosure is better suited to confirming that it continues to advance its AI strategy than to overstating how much its AI revenue has landed. Research conclusion: medium beneficiary / needs verification of renewal rates and AI-course monetization.
Fenbi (HKEX: 2469) Fenbi's investment value lies in professional-exam question banks, mock exams, and civil-service user data, which are inherently highly structured and outcome-oriented and very suitable for AI-enabling. What is truly worth tracking is whether AI raises drilling retention, raises the price per customer for mock-exam/interview services, and forms institutional-side or vocational-education-side B2B revenue. Current public AI financial breakdown is limited. Research conclusion: worth further study / relatively strong content and data moat / financial evidence still needs to be filled in.
Benesse (TSE: 9783) Benesse has long-term accumulation in family education, school education, and publishing content in Japan, so it is theoretically an important beneficiary of AI education, but this round of public materials lacks sufficiently new quantitative disclosure on AI commercialization. Research conclusion: platform-type potential stock / data needs further verification.
"Pick-and-shovel sellers" beyond Microsoft and Google, such as Amazon, NVIDIA, Adobe, Salesforce, and ServiceNow These companies will all benefit in enterprise training, knowledge work, and education-content production, but they lean more toward the spillover of "AI infrastructure, office suites, and enterprise knowledge workflows" than an independent profit pool in the education industry. Research conclusion: indirect beneficiaries / not a core pricing factor for education AI.
Valuation, Risk, and Final Conclusions
Which companies already fully reflect AI-education expectations
Duolingo, Microsoft, and Google have to a greater degree already priced in the AI narrative. Duolingo has the strongest growth and AI brand, but the market also clearly treats it as a core AI-education asset; the education-AI value at Microsoft and Google is real, but it is not the main driver within their overall market caps.
Which companies may have expectation gaps
Platform-type assets like Pearson, Docebo, iFlytek, and Turnitin/PowerSchool/Instructure are more likely to be undervalued by the market as "traditional education / traditional software," while their true value lies in this: what is scarcest in the AI era is not the model, but trustworthy content, master-data systems, school/enterprise procurement relationships, and outcome-verification systems.
Which companies have a "strong AI narrative but insufficient financial verification"
Offcn Education, some Chinese exam-prep companies, several free teacher copilots, and several AI learning-companionship products are more likely to stay at the stage of launching features, pilot partnerships, and user trials, without reaching the latter two steps of "revenue landing—outcome verification—scaled adoption."
Which companies have the greatest near-term earnings elasticity
In the near term, Duolingo, Udemy, Docebo, iFlytek, and Pearson have the greatest earnings elasticity. They either already have AI subscriptions and ARR, or AI is directly driving structural price increases, user growth, or margin improvement.
Which companies have the strongest long-term moats
The strongest long-term moats are more likely to belong to: Pearson (authoritative content + assessment), PowerSchool (school master data), Instructure/Blackboard (LMS workflow), Turnitin (integrity database and institutional relationships), and iFlytek (China government-education channels + hardware + local models).
Which traditional education models are most easily reconstructed by AI
What gets reconstructed first is not high-end education, but: low-end tutoring, general homework answers, static question banks, standardized low-differentiation content, low-threshold content production, and some remote teaching assistance. Chegg has already proven this; 2U's predicament exposes how the pure "channel-reselling + high-cost acquisition" online-education model is more fragile in the AI era.
The five sub-sectors most worth watching
AI language learning: real subscriptions, global reach, strong willingness to pay.
AI enterprise training: enterprise AI reskilling already corresponds to new budgets.
AI exam prep and grading: easiest to charge by outcome.
AI learning management systems and school infrastructure software: strongest channel and switching costs.
AI educational publishing and assessment: authoritative content and assessment rules form long-term moats.
List of listed companies most worth deep research
Duolingo, Pearson, Udemy, Docebo, iFlytek, Coursera, Chegg, Microsoft, Alphabet, TAL. If we start solely from "direct AI-education exposure + verifiable financial elasticity," the first tier is Duolingo, Pearson, Udemy, Docebo, and iFlytek.
List of private companies most worth tracking
Turnitin, PowerSchool, Instructure, Khan Academy/Khanmigo, Speak, Preply, ELSA, Brainly, Quizlet, Workera. These companies cover the four most important secondary profit pools: school main entry points, academic integrity, high-frequency consumer payment, and enterprise skills assessment.
The five points most easily misunderstood by the market
"Having AI features" does not equal "having AI revenue." Many companies are still at the product-launch or free-trial stage.
What is most valuable in education AI is often not the model but content, assessment, and channels.
The school side is not the largest early-mover market; the enterprise side often forms budgets faster.
AI tutors look more like supplementary tutoring and will not fully replace human teachers in the near term.
A low-valuation company is not necessarily an AI beneficiary; it may also be a distressed stock badly hurt by AI. Chegg is the classic example.
Indicators most worth tracking over the next 6-12 months
AI paying-user counts, AI add-on-pack penetration, ARPU changes.
Enterprise learning ARR, NDR, net new customers, AI-course learning minutes.
School/university procurement cases, site licensing, renewal and module-expansion rates.
Learning-outcome research: whether it lifts pass rates, active reading rates, and mastery.
Security and regulation: FERPA/COPPA/GDPR local compliance, minor protection, data-breach incidents.
Risk Analysis
The biggest systemic risk in AI education is not that the model is not strong enough, but commercialization below expectations + insufficient outcome verification + data-security/regulatory backlash. UK research shows that by 2025, 92% of university students were using generative AI, but only about one-third had received formal training; this means campus AI use is expanding quickly while institutions, assessment, and training have not kept pace. Guidance from Japan's MEXT and from public-sector bodies in the UK and the US also emphasizes that schools should use generative AI in a limited and prudent way under conditions of human supervision, safety, and information protection. The PowerSchool breach and extortion incident further shows that if education software cannot handle student data and minor privacy well, it instead loses its most core asset—channel credibility.
Open Questions and Limitations
This research tries to use the latest public materials as of May 19, 2026, but three types of disclosure remain insufficient: First, some Chinese education companies have not yet fully disclosed standalone AI revenue, paying users, and retention; Second, some private companies disclose only financing, customers, and products, not ARR; Third, rigorous randomized trials on "AI improving learning outcomes" remain insufficient industry-wide, with more evidence still resting on product data, usage behavior, and local pilots rather than long-term, cross-disciplinary, extrapolatable outcome verification.
Final Conclusion
AI education is not the infrastructure track that erupts first in the AI value chain, but it is the most typical downstream industry where "model capability must be monetized through content, data, software, and compliance." From an investment-research standpoint, what truly deserves attention is not "who tells the best AI-education story," but:
Who has already taken real money: Duolingo, Pearson, Udemy, Docebo, iFlytek.
Who controls the platform and master data: PowerSchool, Instructure, Turnitin, Blackboard.
Who are the AI-native challengers: Speak, Preply, Khanmigo, Quizlet, Brainly, ELSA, Workera.
Who are the pick-and-shovel sellers: OpenAI, Anthropic, Google, Microsoft, cloud vendors.
Who has already been disrupted by AI: Chegg, low-end answer banks, low-moat course platforms, and some OPM/low-differentiation online-education models.
If we narrow the follow-up research further, the directions most worth prioritizing are: AI tutor, AI language learning, AI exam prep, AI enterprise training, and AI learning management system. These five lines come closest to the intersection of "real revenue, real profit, real retention, and real moats."
This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Full report
Sign in to read the full report
Sign up free to unlock the full text, the Baillie growth scorecard, and full-text search.
Log in / Sign up free