The highest-stakes decisions in K-12 are increasingly being made by algorithms. Accountability ratings. School turnaround triggers. Promotion and graduation outcomes. State assessments now sit at the center of all of it, and the scoring engine is moving from human to AI.
The humans most likely to catch a scoring error in real time, classroom teachers, are walking out of teacher prep programs without any training on AI. New research from the Center on Reinventing Public Education found that when district leaders were asked who's preparing teachers to use AI, not one of them mentioned a school of education.
The deployment is racing ahead. The training pipeline isn't moving.
Some folks have reached out to us asking how to subscribe themselves or their colleagues to this newsletter. Click here to do so. We send a newsletter every two weeks about the latest happenings with K-12 and AI.
IN THIS ISSUE:
What New Jersey's AI-graded state test rollout means for accountability decisions in your district
What CRPE's research says about who's actually preparing the teachers expected to catch AI errors
$58.7M
Two-year contract awarded to Cambium Assessment to develop New Jersey's new NJSLA- and NJGPA-Adaptive tests, with ~75% of essay responses scored by machine. Source
~1,400
Approximate number of Massachusetts MCAS essays incorrectly scored by Cognia's AI in 2025, caught by a third-grade teacher at Reilly Elementary in Lowell. Source
0 of 15
District leaders interviewed by CRPE who mentioned teacher prep programs as a source of AI training. Source
+23 pts
Change in share of CRPE-tracked Early Adopter districts offering AI professional development this year versus last year (86% vs 63%). Source
New Jersey administered its first AI-graded statewide assessments from April 27 through May 29, 2026. The new NJSLA-Adaptive tests were developed by Cambium Assessment under a $58.7 million, two-year contract, with an automated scoring engine handling the essays — Cambium expects only about 25 percent of responses to route to human scorers, leaving roughly 75 percent to the machine. Measurement Incorporated handles flagged "hand scoring." Cambium projected that only about 25 percent of responses would route to humans.
NJ Education Department spokesperson Michael Yaple has noted automated scoring isn't new — last year about 90 percent of essays on the NJSLA and the state's exit exams were already scored solely by machine. That's true. What's new is the adaptive format, the stakes riding on it, and the consequence of getting it wrong.
The cautionary tale is just up the coast. In Massachusetts, Cognia's AI incorrectly scored roughly 1,400 essays on the MCAS. The error was caught by a third-grade teacher at Reilly Elementary in Lowell, not by the contractor and not by the state. Lowell Assistant Superintendent Wendy Crocker-Roberge personally reviewed nearly 1,000 essays and found the AI was deducting points without proper reasoning, including docking students who quoted the passage without using quotation marks.
When state tests are AI-graded, district accountability ratings, school turnaround triggers, and individual student promotion and graduation decisions all rest on a vendor algorithm. The Massachusetts case is not an edge story. It's the evidence of what happens when the scoring engine outruns the audit infrastructure.
OUR TWO CENTS
Districts cannot stop their states from using AI to score state tests. That decision happens above the district level. But districts can get ahead of the consequences.
If parents find out from a news story that an AI graded their kid's test, you've lost the conversation.
This is governance work, not technology work. The question is whether your district has the infrastructure to catch it when AI grades something incorrectly.
Request your state's published handscoring rate for any AI-scored assessment your students take. If they don't publish one, that's your answer.
Build a teacher-facing channel for flagging suspected scoring errors before scores are finalized. Make it short, clear, and route directly to a named person.
Draft a parent communication that names AI's role in your state's scoring before scores arrive. Lead with the audit process, not the technology.
New research from CRPE, asked a simple question: who is responsible for preparing teachers to use AI? In “AI Is Evolving, but Teacher Prep Is Lagging” (Steven Weiner and Robin Lake), researchers interviewed 15 district administrators. None of them mentioned teacher prep programs. Not one.
The implication is significant. Schools of education are the formal pipeline through which the country produces teachers. They set the floor on what a new teacher knows when they walk into a classroom. On AI, that floor is empty. CRPE's broader Early Adopter database shows the share of districts offering teacher AI professional development jumped from 63 percent to 86 percent in a single year. That increase is not coming from higher education. It's coming from districts.
This is the inversion. The slowest part of the K-12 talent system, teacher preparation, is sitting out the most consequential technology shift in a generation. The fastest-moving part, district PD, is filling the gap.
OUR TWO CENTS
The CRPE finding should change how district leaders think about partnerships. If schools of education are not part of the AI training pipeline, the district cannot wait for that to change. The district has to build the capacity itself or partner with an organization that already has.
That doesn't mean writing off colleges of education. It means treating them as a future partner, not a current source. The teachers walking into your buildings this fall will not have AI training from their prep program. The teachers walking in next fall probably won't either. Plan around it.
The other implication is harder. If your district is in a state like Idaho — which just enacted a law (SB 1227) requiring the state to issue AI-use guidelines and every district to adopt an AI policy — the gap between what new teachers can do and what the state expects of them will widen. Districts will be the bridge. Build the bridge now.
Audit what AI training your newest hires actually arrived with. If the answer is "none," your onboarding pipeline is the new prep program.
Build a multi-year AI PD plan that assumes higher education won't help. Internal capacity, district fellowships, peer-led cohorts, partnerships with non-university providers.
Brief your school board on the CRPE finding. The political case for district AI PD investment just got stronger.
The Algorithm: How AI Decides Who Gets Hired, Monitored, Promoted, and Fired and Why We Need to Fight Back Now by Hilke Schellmann
Schellmann is an Emmy-winning investigative reporter at NYU and a Wall Street Journal contributor. The Algorithm is about AI in workplace decisions, but the parallels to AI-graded state tests are direct. She documents what happens when high-stakes algorithmic decisions go wrong, who catches the errors, and how rarely the people most affected ever find out. For superintendents and CIOs considering audit infrastructure for AI-scored assessments, this is a prescient read.
Chat with ClassCloud
We’re listening. Let’s Talk!
This newsletter works best when it’s a conversation, not a broadcast. If you want to talk through how any of this applies to your district specifically—or if you have feedback on what would make this more helpful—just hit reply. We read and respond to everything.
Schedule a Virtual Meeting
Thanks for reading,
Russ Davis, Founder & CEO, ClassCloud ([email protected])
Sarah Gardner, VP of Partnerships, ClassCloud ([email protected])
ClassCloud is an AI company, so naturally, we use AI to polish up our content.
