AI Is Coming to Education. The Teacher Isn’t Going Anywhere.

Education is a public commitment before it is a market. It is a massive industry, with over a trillion dollars spent annually on both K-12 and workforce education. But it has also bedeviled technological disruptors – established players have maintained market dominance, while well-funded upstarts (Knewton, AltSchool) have failed to deliver against their promise.

Education is changing. AI is coming, regardless if stakeholders want it. But teacher shortages, changing funding models, and a push for workforce development programs are also creating new challenges and opportunities for schools, employers, students – and innovators.

Technologies that can harness these trends and prove they can deliver measurable results – to students and teachers, parents and administrators, even employers – have an opportunity to win a meaningful piece of and even grow that market.

But innovators need to understand what makes education unique, as a social commitment and as a market. Educators have strong preferences and low tolerance for incomplete solutions. Buying decisions are made across many stakeholders in highly structured ways. Students, teachers, and parents are all highly skeptical of AI. And many of the most important challenges in the classroom cannot be solved with technology.

Startups that can navigate these constraints and capitalize on these trends are positioned to finally, successfully, transform American education and grow the ed tech market.

Education is a trillion-dollar problem, but not a trillion-dollar software problem. The US spends over $980 billion a year on public K-12, and an additional $100 billion on workforce development. However software only represents 0.5% of K-12 spend, and has flatlined in recent years. It’s an important industry, but many of its challenges cannot be solved (at least not well) by software.

Public school spending Nationally, American school districts spend about $980 billion a year, roughly $16,500 per student. However most of that spend is not for software – for example publishers sold $5.2 billion of instructional materials into K-12 in 2025, a slight decrease from 2024. That said, software has become a bigger part of education since the COVID pandemic. The average district has access to around 3,000 different software tools.

Workforce development Increasingly, public schools are investing in training students for careers of the future. In recent years, schools spent $7.3 billion annually on Career and Technical Education (CTE), including out of classroom tools and programs.

But education isn’t just for kids. Employers spent $102.8 billion training their own people in 2024–25, up about 5%, or $874 per learner. This area is becoming more important as businesses and employees work to retrain for the AI economy.

Tutoring Beyond formal education, families spend heavily on out of school tutoring. While data is limited, the two largest public companies in US tutoring and test prep – Kaplan’s supplemental education arm and Varsity Tutors – together do about $500 million a year. However market researchers estimate total spend is in the tens of billions when including the informal tutoring economy. This is a market of individual tutors, franchise centers and cash, with no scaled incumbent.

Ken Koedinger, developer of the Cognitive Tutor, said that “step change is what 25 years of incremental change looks like from a distance.” Change here is incremental. But specific developments in technology, workforce composition, and funding create unique opportunities for ed tech in the coming decade.

AI is here Most obviously, AI has created a technological inflection point across industries, including education. Already, generative AI has transformed both how educators create curriculum and how students do homework – sometimes at the expense of learning.

Deterministic computing used to constrain educators’ use of tech to subjects with discrete problems and explicit answers. Generative AI, trained with pedagogical expertise and enriched with personalized student data, could allow educators to easily create dynamic, customized tooling to support students.

This isn’t just theoretical. In NJ, a HS teacher used Playlab’s platform to build a bespoke chatbot that gave real-time feedback on student revisions to an instructor-graded draft. His students got roughly ten times more writing feedback in the first year he used it – more importantly, they saw a measurable improvement on their AP test performance.

Teacher shortages There is a growing talent gap. In 2024–25, about 425,000 teaching positions (roughly one in eight) were either unfilled or filled by someone not certified for what they were teaching. Shortages were widest in special education, science, and math, as well as in the burgeoning field of career and technical education. This growing need for qualified educators has created specific opportunities for technology to help fill the gap – it can enhance teacher capacity, and even provide bespoke technical education.

Funding changes Policy is changing. New sources of public and private funding are creating opportunity for innovators without requiring them to disrupt entrenched incumbents or navigate school district procurement processes to access dollars.

Federal Perkins grants and state initiatives are funding Career and Technical Education directly: states invest nearly $6 billion in CTE, on top of $1.44 billion of federal grants. And Congress recently introduced Workforce Pell grants – they’re still small (roughly $30 million a year), but are tied directly to completion rates, employment and graduate earnings.

Programs like Education Savings Accounts and the Federal Scholarship Tax Credit also create avenues for public and private money to reach private schools and create opportunities for tools to reach students outside of established school district procurement processes. Education Savings Accounts already reach about 489,000 students; Texas alone appropriated $1 billion for its program starting in fall 2026, at an average award of $10,500. Separately, the Federal Scholarship Tax Credit (starting in January 2027) provides substantial tax incentives to donate to private scholarship organizations.

There is no single problem for technology to solve in education. Tools need to prove to stakeholders – parents, teachers, students, or administrators – that they measurably improve performance. In a complex ecosystem, the solutions that demonstrably help buyers accomplish their goals (improved test scores, saved administrative time, increased student engagement) are most likely to succeed.

Proof matters, but what counts varies Education has no single agreed measure of what “working” means – the goals of schooling are contested and plural, while political and social emphases swing between test-score accountability and whole-child measures.

Most vendors have research showing their products “work” but vary in how they define success. The tools most likely to break out need to take the specific buyer objective and prove against it.

School Districts need to meet state metrics While it varies across states, school districts are generally evaluated against several performance metrics, including test scores, attendance, and graduation rates. School districts have explicit financial and administrative incentives to improve and maintain these numbers – they already invest substantial resources, in consultants, headcount, and software, to address these metrics.

Increasingly, school districts are also investing in career and technical education for students. Many states have allocated money to career programs but require proof of career outcomes (a livable wage, a high-growth field, real demand for the skill) to be eligible.

Students and families are focused on scores and credentials Students and parents are motivated by standardized tests, credentials, and, eventually, degrees that unlock new opportunities and employment, particularly as students advance in their education. Tutoring and test prep are already well-established industries, but they are fragmented and built on legacy technologies. They are ripe for disruption.

Teachers want their time and focus back Teachers want time back from administrative tasks, but instead are increasingly asked to expand their reach to fill the teacher gap. There are myriad responsibilities, including administrative tasks (recording grades, scheduling homework), and lesson plan customization, that pull teachers from focusing on student instruction. They also face ongoing pressure to improve student performance. Teachers will embrace tools that save them time and automate workflows, help them meet administrative requirements, and improve student outcomes, and anything else that gives them more time to focus on instruction.

Engagement, but outside the classroom Engagement is not the same as learning – a product can win attention without teaching anything. Critically, engaging content cannot replace the student-teacher relationship. Students want a relationship with a teacher inside the classroom.

But that pull is much weaker for homework and after-school work, where a game or a bot competes with non-educational distractions. Beyond explicitly academic goals, there are opportunities for tools that get students excited about learning. The most successful of the ed tech platforms, like Duolingo, embrace gamification and highly engaging content to increase the frequency and duration of student use.

Education technology is littered with innovators who failed to deliver against the hype and funding. While recent technologies and trends address some of these failures, other issues were structural. Education is a relationship business, run by cautious institutions, sold through a fragmented buyer, and serving a public that is increasingly wary of AI itself. Successful tools can’t just “sprinkle AI” onto education – they need to understand what motivates students, teachers, parents, and administrators, and recognize what parts of education relationships are inviolable.

Students want to learn from teachers Students care about what their teachers (and people generally) think – technology cannot care the way a teacher does. Justin Reich’s research locates motivation in that relationship: most of what motivates a student to learn is their social connection to the people around them. Unsurprisingly, students repeatedly report wanting feedback from a human being, the way a patient prefers talking to a doctor, or a caller prefers a live agent.

Teachers value this as well. They know students learn best from teachers they respect and like, and they rely on their relationships with students to identify when they have stopped engaging.

A product that removes the teacher from the loop or weakens that relationship will fail. An AI companion can complement learning goals outside of the classroom, but it cannot stand in for a person in the room.

K-5 classrooms particularly need teachers In the earliest grades, teachers goals are only partly academic. Teachers spend a large part of primary education helping young kids learn to sit still, pay attention, work with others, and learn to learn.

Software is good at supporting measurable, academic goals, but it struggles with the myriad and complex goals of the youngest classrooms. Today, teachers describe young students half-listening and clicking through programs if they have no adult beside them. Even a well-run one-to-one deployment does not engage every child, and the time spent moving on and off devices eats into the instructional block.

Conservative users demand polish and support Teachers are cautious to adopt new technologies that ignore decades of pedagogy. They have watched underbaked tools arrive and fail, with limited training and support.

Build fast and break things doesn’t work for education. Hands-on training and support does. Educators trust other educators – the human layer is what buys tolerance for software that is not yet perfect. Even well-designed and tested tools will fail, if teachers aren’t supported with guidance. Teachers won’t completely change how they teach to accommodate new tools – they will attempt to insert them into their existing paradigm, if they use them at all.

A complicated buyer market Public education systems are difficult to sell into. These institutions buy a lot of software, but the decision is split across parties: the state sets standards and allocates funds, the district signs the contract, teachers have to decide how to use tools, students and parents experience the impacts. A product has to be legible to all of them at once. The risk is misreading who chooses, who implements, and who pays.

There is an operational version of the same risk that surfaces in conversation with teachers: once the school year begins, there isn’t time or appetite to try new things – there is effectively one buying window a year, which shapes the cost of reaching a customer.

AI backlash In the corporate world, technologists have used AI hype as a marketing tool. There’s an opposite dynamic in education. Parents are instinctively defensive of student privacy and performance. Teachers feel like AI tools have been non-consensually thrust upon them and their students, and have undermined many of their educational and evaluation tools.

Most stakeholders are not clamoring for more AI in education. Solutions must help people accomplish existing, explicit goals, even when they may otherwise be skeptical of the technology.