Critics often misunderstand the transformative potential of educational technology. A recent Wired article questions the promise of artificial intelligence in schools, pointing to recent trouble at Alpha Schools—the most prominent and controversial “AI charter school”—as a case study. Critics and parents featured in the piece contend that the model, which replaces traditional teachers with non-instructional “guides,” places excessive pressure on students to meet software metrics, resulting in anxiety, weight loss, and self-harm among children.
While the article points out real problems with Alpha Schools, it also advances a common misconception: that AI education consists primarily of foundational skills practice through platforms like IXL. In reality, the K-12 innovation landscape is vast, spanning adaptive practice, intelligent tutoring, teacher-facing analytics, and generative copilots. Mistaking a discrete component of the modern classroom for the entirety of AI innovation obscures how personalized learning can empower educators and close the achievement gap for the nation’s most vulnerable students.
This isn’t simply academic for me. I founded and led Kairos Academies, a personalized-learning charter network that used the Summit Learning Platform and integrated tools like IXL. I have seen both the advantages and the drawbacks of these technologies in real classrooms.
Why skill drills matter
Children learn multiplication tables through repetition. Rote memorization that 3 x 4 equals 12 is not a distraction from “real math”; it’s a prerequisite for it.
Decades of research show that conceptual understanding and procedural fluency develop iteratively and mutually reinforce each other. Students cannot reason flexibly about functions if their working memory is consumed by basic operations. Cognitive load theory explains why: when the brain automates lower-level steps, it frees scarce working memory for higher-order reasoning.
This is exactly why practice—whether on paper or on a computer—matters for advanced problem solving. Modern software can help. A large body of evidence on intelligent tutoring systems, such as IXL, finds consistent learning gains over teacher-led direct instruction or static worksheets. Meta-analyses report positive effects across subjects and grade levels, with especially strong benefits for low-income students who start behind their peers.
These systems do not replace teachers; they offload the repetitive, right-or-wrong, feedback-intensive parts of practice so that teachers can spend more time diagnosing misconceptions and coaching higher-order thinking. Randomized evaluations of tutoring and practice platforms in the United States show measurable improvements when students receive immediate feedback and when teachers use the resulting analytics to target instruction. Findings include higher math performance versus traditional homework and improved assignment completion when teachers engage with reports. While headlines often herald these tools as “AI replacing teachers,” the software actually augments foundational practice, making practice data visible and actionable so teachers can focus on high-leverage instruction.
Personalized learning and school choice
Personalized learning is not a fad invented by Alpha Schools, which is designed around its own “2 Hour Learning” model. The school claims that students master core academics in just two hours a day using “personalized, AI-driven software” instead of traditional teacher-led instruction. Adult “guides” supervise the classrooms and manage student motivation, freeing up the rest of the day for life-skills workshops and entrepreneurial activities. Proponents claim this adaptive, technology-first approach allows students to learn at twice the speed of their peers in conventional schools.
Whether or not Alpha Schools’ particular model succeeds, prior multi-site studies of personalized learning, such as the RAND Corporation’s “Continued Progress” report, have demonstrated the success of technology-driven self-directed education. Personalized learning schools, such as those that used predecessors to AI tools, have documented positive gains, better achievement monitoring, and higher student awareness of goals and progress.
By contrast, teacher-controlled direct instruction aims at the “average” student, leaving slower students confused while hindering growth for faster peers in the class. AI in education, which will take a multitude of forms, holds particular promise for low-income students. They have idiosyncratic needs given disproportionate special education diagnoses, inconsistent attendance, and understaffed schools. Their range of mastery is also wider than wealthier peers; students who are not grade-level proficient may be any number of grades behind. The kind of personalized learning that AI unlocks would be particularly transformative for America’s most underserved students.
That said, critics rightly flag concerns regarding Alpha Schools’ culture and incentives. For example, students should never need to “earn” food, and accommodations should not hinge on rewards. Federal health guidance rightly discourages using food as a reward precisely because it undermines healthy norms and stigmatizes students. More broadly, decades of experiments have shown that extrinsic rewards can crowd out intrinsic motivation when misapplied. Personalized learning is about agency, relevance, and mastery, not compliance for prizes.
Criticisms of this particular implementation should not lead policymakers to dismiss other AI education models that foster student autonomy, flexible pacing, and modern work environments. Giving adolescents structured independence—backed by clear mastery checkpoints and adult coaching—prepares them for the postsecondary reality where self-management is non-negotiable. Technology like the platforms Alpha Schools is testing will eventually automate everything rote in classrooms, letting students move at their own pace, revisit gaps immediately, and learn with multiple modalities (text, video, and simulations). It will free time for teachers to do what they do best: subjective judgment and value-based decisions around curriculum, evaluation, and human connection.
The Wired piece points to Alpha Schools’ attrition as an indictment of AI in education. To the contrary, family choice and even school closures are features of healthy school markets. Market dynamics—including right of exit and the potential for failure—show why a pluralistic public system is necessary, particularly with emerging technologies. In a brand new field like AI education, many competitors will try different models, and some of them will underperform. The charter compact empowers parents to exit underperforming schools; those schools then lose the public funding attached to those students, forcing accountability and preventing them from continuing to fail students. Students are naturally directed towards higher-performing schools that remain. Even if parents leave Alpha Schools, that doesn’t mean they won’t love a different AI education model.
The urgent need for innovation
AI has arrived at a pivotal moment for American education. Looking at international benchmarks, math performance in the United States declined in Programme for International Student Assessment 2022 and sits around the OECD average. That means American teenagers are less prepared to apply math in real-world settings than those of about 40 other developed nations. Reading and science are better, but not world-leading. National Assessment of Educational Progress results through the 2024–2025 school year show broad declines since 2019, especially for lower-performing students.
Today’s education system doesn’t work for most students, especially those in low-income communities. Policymakers and educators must be urgently trying new approaches, and AI education is the most exciting and promising breakthrough since Chromebooks became ubiquitous in schools ten years ago.
Here are some improvements to AI-enabled models:
- Maintain basic standards of culture and care
No food-for-performance policies; honor special needs accommodations. Pair analytics with humane, relationship-rich advisories and frequent teacher-student conferences. - Procedural fluency plus rich tasks
Use adaptive practice to build automaticity, then spend protected class time on problems that require explanation, modeling, and argumentation. That is how teachers reduce cognitive load while raising rigor. - Real-time, transparent accountability
Require mastery-based checkpoints with frequent, low-stakes assessments. Use platform telemetry to surface engagement and misconceptions to teachers and parents. Publish comparable growth metrics across schools and sectors. Learning-analytics research and tutoring platforms show how to convert “big data” into actionable next steps. - Keep humans at the center
Use AI to draft differentiated materials, generate additional practice at the right level, and summarize class trends. Use teacher time for professional, subjective judgments. Educators should be asking: Why is a student not understanding this concept? What feedback will unlock it? How do I connect this work with a child’s life goals?
Advocates of school choice should welcome AI-enabled models as a vital component of a pluralistic system. Implementation challenges at a single school do not impugn the technology itself; rather, they highlight the importance of market competition in refining how we serve students. In an era of stagnant achievement and post-pandemic declines, policymakers cannot afford to block innovations that personalize learning.
AI offers the power to give every child a one-on-one tutor, tailored to their specific pace, knowledge gaps, and learning style. That helps everyone, especially low-income students who have historically received the least individualized attention. Educators should embrace this technology, leveraging it where it excels so that teachers can focus on uniquely human mentorship and instruction. Authorizing more AI charter schools will empower families to choose among increasingly diverse, personalized, and accountable options.