S17.11 · Education, Training & Human Capital
An $85B market bifurcated between AI-native consumer apps and slower-growing institutional instruction.
Language learning covers instruction and software teaching foreign-language proficiency, sized at $85.1B (2025, Global Market Insights). The headline forecast — a 22.9% CAGR to $649B by 2035 — deserves caution: it extrapolates consumer-app subscription growth across a category whose institutional classroom half is running much slower, so it should not be read as a blended forecast for the whole. The structural event sits elsewhere in any case. Agentic AI is collapsing the cost of live conversational instruction specifically, because AI conversational partners now substitute for human instructors on the two dimensions live schools have always sold: price and availability.
One nominal category, two growth profiles. The consumer-app segment, led by Duolingo and its peers, is compounding double-digit; institutional and classroom instruction is growing low-single-digit. Revenue concentrates in the US and Europe on the strength of app subscription spend, while Asia concentrates the institutional English-instruction demand — China, South Korea and Japan's eikaiwa conversation-school market. Ownership matches the bifurcation: the app layer is concentrated, with Duolingo dominant and Babbel and Busuu beside it, while in-person instruction fragments across thousands of local schools and corporate-language vendors.
Content and curriculum providers and native-speaker instructor supply sit upstream. Downstream, individual consumers are the largest payer, with employers funding global-mobility training and government immigration-integration programs beside them. The two halves earn differently. The app layer runs SaaS-like economics — freemium-to-subscription conversion, high gross margin, durability tied to subscriber and daily-active-user retention. In-person instruction earns a billable-hour spread, and margin lives or dies on it. Regulatory gating is light overall, limited mainly to work-visa and citizenship language-certification requirements — but where it applies, it creates a genuinely captive demand pocket that neither an app nor an AI substitute can currently satisfy on its own.
The adjacencies run in three directions: EdTech platforms supply infrastructure, tutoring overlaps directly wherever live instruction crosses between the two categories, and corporate learning connects through employer-sponsored global-mobility training programs. The product moves sort by what they do to the revenue line. Adding live-tutor and certification features protects app subscription revenue against pure-AI substitutes that could otherwise be had for free — a defense of existing dollars. Extending into adjacent gamified-app categories — math or coding, for instance — adds new ones, applying a proven consumer engagement model to fresh subject areas.
The AI displacement is precise in its target: the cost of live conversational instruction is collapsing because AI conversational partners now match human instructors on price and availability, a direct threat to in-person and live-tutor language schools that have historically competed on exactly those two dimensions. What holds is what regulation and physical presence protect — certified-proficiency testing tied to visa and citizenship requirements, where the exam itself must carry a regulatory recognition an AI interaction cannot confer, and immersion or cultural-context delivery that a conversational AI tool does not replicate.
AI conversational practice is already underway and mainstream among app users. Near-total substitution for casual-learner live instruction is estimated within 2-5 years, constrained specifically at the certified-proficiency-exam layer, where regulatory recognition requirements move on a much slower clock than the underlying technology.