Districts Are Buying AI They Cannot Evaluate

Los Angeles Unified paid roughly $3 million for an AI chatbot. The vendor, AllHere, later filed for bankruptcy. That is the version of the story that travels, because failure with a dollar sign attached is easy to narrate. But the quieter figure in Robbie Sequeira's August 12 report for Stateline is the one school leaders should sit with: districts have access to an average of 3,001 digital tools, and educators and students actually use about four.

Read those two numbers together and a different problem comes into focus. Districts are not failing to buy AI. They are buying AI faster than they can judge it, and the tools that do not survive first contact with a classroom simply go quiet rather than getting cancelled.

The Asymmetry Is the Real Story

Mark Schneider of the American Enterprise Institute names the structural issue plainly in the Stateline piece: "There's always an asymmetry between what the providers know and what districts know." That asymmetry has always existed in education technology. AI widens it, because the product changes between the demo and the deployment, and because the vocabulary is new enough that a procurement committee cannot easily tell a strong claim from a decorated one.

The market is scaling into that gap. AI in education generated about $2.5 billion in the United States in 2024 and is projected to pass $15 billion by 2033, inside a broader edtech market moving from roughly $48 billion toward $90 billion by 2030. Money is arriving well ahead of any shared standard for judging what it purchased.

The Burden Sits in the Wrong Place

Carol Birks, superintendent of Pennsylvania's Allentown School District, identifies where the weight lands: "A significant challenge is that the onus of making these decisions is placed entirely back on the individual school district." Scott Langford, superintendent of Sumner County Schools in Tennessee, is blunter about the yield: "Most of them don't really do much that drives any kind of change for students."

Some states have started to help. Pennsylvania guidance directs districts to understand data control, examine third-party practices, limit collection, and maintain human oversight for grading and discipline. New York City folded AI standards into its privacy review and prohibited the use of student data for model training. Chicago published an AI guidebook and began blocking unapproved third-party AI products at the network level. The Southern Regional Education Board issued a procurement and evaluation checklist. This is real progress, and it is still thinner than the purchasing volume it is meant to govern. Leslie Eaves of SREB compresses the current mood into five words: "Anything with AI is a risky click."

Five Questions That Separate Instrumented AI From Impressive AI

Dacia Toll of Coursemojo notes that "education has had a long-standing issue with vetting edtech for safety, usability and efficacy." AI did not create that gap, but it does punish it faster. These are the questions we would want any district to put to us, and to every other vendor in the room.

Pedagogy Belongs on the Procurement Rubric

There is a design distinction hiding inside the efficacy question. An AI that hands students answers generates activity data and very little else. An AI built on Socratic questioning keeps students in question space, and the record it produces is a record of reasoning — which is the only thing a district can actually evaluate against a standard. That is why we built our platform around continuous mastery tracking rather than usage counts, and why the tutor is designed to adjust reading level while holding academic rigor constant across 150+ languages.

None of that is a claim of outcomes. It is a claim about what kind of tool can be held accountable for outcomes at all.

Pilot Before Portfolio

The cheapest correction available to a district right now is sequencing. Measure on two courses before scaling to twenty thousand students. Define what would count as success before the vendor defines it for you. Insist that the data flows back to you in a form your assessment team already knows how to read. That sequence is exactly why our research pilot exists: partner institutions deploy on two courses, keep full platform access at no cost during the pilot, and share anonymized usage and outcome data so the results can be published rather than asserted. We are studying whether this works. That is a different posture than selling certainty, and it is the posture the current market badly needs more of.

Districts do not need to spend less on AI. They need to stop buying anything that cannot be graded.

Related reading on our approach: the research pilot program and our certification alignment coverage.

Read the Full Article

Read "Schools spend billions on AI, but struggle to figure out what's worth buying" on Stateline

About Core Learning Exchange: We provide turnkey Career and Technical Education (CTE) solutions for grades 6-14, offering 450+ courses from 20+ providers aligned to state standards and industry certifications. Our AI platform uses proven Socratic methodology to develop critical thinking skills through personalized, adaptive learning—deployed in hours via LTI integration.

Related Posts