The Changemaker — August 2026
Four things that happened in the field this month, and what each one actually teaches.
AI is drafting IEPs faster than districts are writing policy for it
A national survey published by the Center for Democracy and Technology found that most special education teachers used artificial intelligence to help develop an IEP or Section 504 plan during the 2024-25 school year. That number should stop a director mid-scroll. Not because AI has no place in an IEP meeting, but because most of that use is happening without a district policy telling teachers what is and is not acceptable, and without anyone checking whether the output actually meets IDEA's requirements.
By early 2026, thirty-three states had issued some form of AI guidance for educators. That sounds like progress until you notice the gap: guidance is not the same as a workflow, and a workflow is not the same as a review process. A teacher drafting a present level of performance with an AI tool at 9:00 PM the night before a meeting is not violating any law by doing so. The violation risk shows up later, in what that draft actually contains.
The distinction AI tools do not know
Specially designed instruction is a defined legal term, not a teaching strategy a district can swap in and out. Federal regulation defines it as adapting the content, methodology, or delivery of instruction to address the unique needs that result from a student's disability, and to ensure access to the general curriculum (34 CFR § 300.39(b)(3)). An accommodation, like extended time or preferential seating, does not change what or how something is taught. It changes the conditions around the same instruction every other student receives.
A general-purpose AI tool has no way to know which one it is producing. Ask it to write "supports for a student with a reading disability" and it will often return a list that blends both categories freely, accommodations and SDI sitting side by side with no legal distinction drawn between them. If a teacher pastes that list into an IEP without separating the two, the document now describes accommodations as if they were specially designed instruction, or omits SDI the student is legally entitled to because an accommodation was substituted for it without anyone noticing.
Education Week reported in early 2026 that advocates are raising exactly this concern: teachers are increasingly using AI to speed up IEP writing, and the tools are outpacing districts' ability to write policy or train staff on their limits. A brief published by the Center for IDEA Data and Digital Learning (CIDDL) in April 2026 reached a similar conclusion from the research side, noting that generative AI can support the drafting process without diminishing goal quality, but only when a trained educator reviews and edits the output with genuine clinical reasoning, not a rubber stamp.
What "meaningful review" actually requires
None of this makes AI assistance improper. It makes unreviewed AI assistance improper. The Center for Democracy and Technology's own analysis states plainly that an AI tool generating an IEP based on little student-specific information, without significant teacher review, would likely not meet IDEA's requirements. The team's signature is what makes an IEP legally binding, and a signature on a document nobody actually checked is a compliance exposure waiting to surface at the next due process complaint or state monitoring visit.
For a director, the practical questions are not about banning a tool. They are about whether your team can answer three things right now: What AI tools are teachers already using for IEP drafting, whether your district has said anything about that use in writing, and whether anyone is checking that the SDI in a finished IEP is actually SDI, and not an accommodation that got labeled that way by a tool that did not know the difference.
That third question is the one most districts cannot currently answer. It is also the one a state monitoring visit will ask first.
Where the Lab can help
IEP Builder, by The Lab, was built around the assumption that AI assistance in IEP writing is not going away, so the more useful question is how to keep a trained educator's clinical reasoning in the loop while the document gets written, not after. The tool coaches the distinction between SDI and accommodation as a teacher drafts, flags goals that are not measurable before they are finalized, and keeps the compliance calendar visible throughout, rather than treating quality as something to check at the end.
1. Center for Democracy and Technology, cited in Riddell, R. (2025, November 4). Heightened AI use in special education brings elevated risks, K-12 Dive.
2. Education Week. (2026, January 30). Teachers are using AI to help write IEPs. Advocates have concerns.
3. Center for IDEA Data and Digital Learning (CIDDL). (2026, April 6). Generative AI and IEP goal development: Implications for special education teacher preparation.
4. 34 C.F.R. § 300.39(b)(3) (definition of specially designed instruction).
The Supreme Court case behind every "is this goal good enough" conversation
In 2017, the Supreme Court decided Endrew F. v. Douglas County School District RE-1, and special education leadership has been living inside that decision ever since, whether a team names it in the meeting or not. The case reached the Court because a family argued their son's IEP goals had become so repetitive and so modest that the district was, in practice, offering little more than the same plan year after year. The Court agreed the standard had to be higher than that, and it said so in a sentence every director should be able to recite: an IEP must be reasonably calculated to enable a child to make progress appropriate in light of the child's circumstances.
That standard replaced a much lower bar some courts had been applying, one that asked only whether a child received some benefit, however minimal. Endrew F. did not hand districts a numeric target. It handed them a question that has to be answered with evidence: can this team show the progress this standard requires? And that question cannot be answered if the goal underneath it was never written in a way that can be measured in the first place.
The four parts a measurable goal cannot skip
Federal regulation requires that an IEP include a statement of measurable annual goals, and that the IEP describe how progress toward those goals will be measured (34 C.F.R. § 300.320(a)(2)). In practice, a goal that actually meets this standard needs four identifiable parts, all in one place: the condition, given what situation, material, or level of support; the student, named directly, not folded into a passive sentence; the observable behavior, a specific, visible skill, not a feeling or an impression; and the criterion, how well, how often, and measured how.
Miss any one of the four and the goal cannot be evaluated with real data. Texas's state guidance on this point is blunt: even a goal written in seemingly measurable terms, such as "70% mastery of grade-level standards," does not satisfy 34 C.F.R. § 300.320(a)(2)(ii) if it is really just a restatement of what every student in the general curriculum is already expected to do. A goal has to describe this student's individualized target, not the curriculum standard the whole class is working toward.
Why "making progress" is not a progress report
The PROGRESS Center, an OSEP-funded technical assistance center, has published guidance specifically on this failure point in the years since Endrew F. was decided: progress reporting has to use the same measurement method as the goal itself, and it has to produce numbers, not impressions. "Making steady progress" is not evidence. "Read 42 words per minute with 90% accuracy, up from 28 words per minute in September" is evidence, because a reviewer, a parent, or a hearing officer three years from now can look at that sentence and know exactly what happened without asking anyone to interpret it.
This is where the legal standard and the instructional standard become the same standard. A team that cannot produce that second sentence for a given goal has no way to demonstrate the "appropriate progress" Endrew F. requires, regardless of how well-intentioned the instruction actually was. The paperwork is not separate from the compliance exposure. It is the only record of whether the compliance exposure exists.
What this means for a caseload review
Most districts do not find out a goal was never measurable until a parent asks a pointed question in a meeting, or a state monitor pulls the file. A faster, lower-stakes way to find the same gap is to pick ten IEPs at random and ask, for each annual goal, whether someone outside the room could look at the goal and the most recent progress report and know exactly what happened without asking a follow-up question. If the answer is no more than half the time, the exposure is not with one teacher. It is with the review process, or the absence of one.
1. Endrew F. v. Douglas County School District RE-1, 580 U.S. 386 (2017).
2. 34 C.F.R. § 300.320(a)(2) (content of the IEP; measurable annual goals).
3. Texas Education Agency, Special Education Support. Question and Answer Document: IEP Measurable Annual Goals.
4. PROGRESS Center. IEP Tip Sheet: Measuring Progress Toward Annual Goals, U.S. Department of Education, Office of Special Education Programs, Award No. H326C190002.
Caseloads are now a strike issue. The research points past staffing alone.
In San Diego this year, more than 90% of teachers voted to authorize a strike, and the vote was not only about pay. EdSource reported that special education caseloads sat at the center of the dispute, alongside negotiations happening in the same window in San Francisco and West Contra Costa. Districts have started writing caseload relief and special education recruitment stipends directly into negotiated agreements. That is a meaningful shift: special education staffing has moved from a line item in a budget memo to a headline demand at the bargaining table.
The instinct that follows is to treat this purely as a hiring problem. Post more positions, offer a signing bonus, reduce caseload numbers on paper. Those moves matter, and in a genuine staffing crisis they are sometimes the only lever available in the short term. But the research on what actually keeps a special education teacher in the classroom, and effective once they are there, points to something districts control more directly than the labor market: how that teacher is developed once they are hired.
What the coaching research actually says
The most rigorous evidence on this question comes from a 2018 meta-analysis by Matthew Kraft, David Blazar, and Dylan Hogan, published in the Review of Educational Research. Combining results across 60 studies that used causal research designs, not just correlational surveys, the authors found that individualized instructional coaching improved teachers' classroom practice by 0.49 standard deviations and student achievement by 0.18 standard deviations. Those are real, replicated effects, not a single promising pilot.
The same analysis carries an important caveat that gets dropped when the finding is repeated as "coaching works": the effect was strongest in smaller, tightly implemented programs, and it diminished as programs scaled without maintaining the same individualization and sustained contact. A coaching program that becomes a once-a-semester walkthrough with generic feedback is not the program the research describes, even if it uses the word "coaching" in the title. The active ingredients are specificity, individualization, and sustained contact over time, not the label on the initiative.
Why this belongs in the same conversation as caseloads
A teacher who feels unsupported and under-coached will leave a high-caseload position faster than a teacher who has both a manageable caseload and someone showing up regularly to help them get better at the hardest parts of the job. Caseload relief addresses the first half of that sentence. Job-embedded coaching addresses the second half, and it is the half a district can build internal capacity for without waiting on a labor market that is not going to loosen soon. The Bureau of Labor Statistics and multiple state shortage reports agree on this much: special education has been the most commonly reported teacher shortage area in the country for years, and there is no credible projection where that reverses on its own in the next contract cycle.
Treating caseload relief and coaching investment as two separate line items, one for the bargaining table and one for the professional development budget, misses the point the research is actually making. They are the same retention strategy, addressing two different points of failure in the same system.
What this looks like in a budget conversation
A director walking into a budget or negotiation conversation this year has a more defensible position with a specific claim than with a general one. "We are underfunding coaching relative to what causal research shows moves teacher retention" is a harder sentence to dismiss than "our teachers need more support." The first sentence can be checked against a citation. The second cannot be checked against anything, which is exactly why it tends to lose the budget argument.
1. Espinosa, A. (2026, July 23). Special education has become a flash point in negotiations with teachers, EdSource.
2. Kraft, M. A., Blazar, D., & Hogan, D. (2018). The effect of teacher coaching on instruction and achievement: A meta-analysis of the causal evidence. Review of Educational Research, 88(4), 547-588.
3. Learning Policy Institute. (2026). State Teacher Shortages 2026 Factsheet.
A federal appeals court just revived a case over grouping students by IQ
On October 9, 2025, the Tenth Circuit Court of Appeals issued a ruling in Jacobs v. Salt Lake City School District that every team making placement decisions should read, not because it settles the underlying dispute, but because of what it says about how that dispute has to be resolved. Two elementary school students with intellectual disabilities, identified in the case as E.J. and H.S., along with the Disability Law Center, had challenged a district practice of automatically assigning students with intellectual disabilities to self-contained classrooms at a limited number of designated "hub" schools, based on IQ-based categories, without an individualized assessment of whether a general education classroom with supplementary supports might be appropriate for a given student.
A federal district court had dismissed the case, reading the plaintiffs' claims narrowly as a request to attend their neighborhood schools, a form of relief the Tenth Circuit had already held was unavailable under IDEA. The Tenth Circuit disagreed with that reading on appeal. It reversed the dismissal and sent the case back to the lower court, holding that the plaintiffs had plausibly alleged something distinct and viable: that the district's hub system, by sorting students into placements based on IQ category alone, failed to make the individualized placement determination that IDEA, the ADA, and Section 504 all require for every student, every time.
What the ruling does and does not establish
It is worth being precise about what this means. The Tenth Circuit has not ruled that Salt Lake City School District actually violated the law. That question goes back to the district court for further proceedings. What the appellate court did establish is narrower and still significant: a district cannot satisfy its individualized-placement obligation by sorting students into a category first and calling that category a placement decision. The claim that a district did exactly that is enough, on its own, to get a case past a motion to dismiss and in front of a court.
The regulation this case is really about
Least restrictive environment gets treated in a lot of IEP meetings as shorthand for a room: which classroom, for how many minutes a day. That is not what the federal regulation actually says. LRE requires that, to the maximum extent appropriate, children with disabilities are educated with children who do not have disabilities, and that removal from the general education environment occurs only when the nature or severity of a disability means education there, even with supplementary aids and services, cannot be achieved satisfactorily (34 C.F.R. §§ 300.114-300.117). The regulation is a statement about instruction and support first. Placement is the conclusion that follows from an individualized answer to that question, not the starting point.
A team that opens a placement conversation with "which room does this disability category usually go to" has already inverted the sequence the regulation requires, whether or not anyone in the room intends to shortcut anything. The Jacobs case is what that inversion looks like at the scale of a district system rather than a single IEP meeting, and it is why a disability rights organization was willing to litigate it as a systemic claim rather than a single family's dispute.
What this means for a placement review
Every district groups students in some way. Grade-level teams, schedules, and staffing models require it. The legal exposure is not the existence of categories. It is a decision process that starts and ends with the category, without a documented, individualized answer to what this specific student needs instructionally and whether the setting under consideration can meet that need with appropriate support. A useful audit question for any placement pattern: if the category were the only factor a hearing officer could see removed from the record, would there still be evidence of an individualized decision underneath it? If the honest answer is no, the pattern carries the same exposure the Tenth Circuit just allowed to proceed to litigation.
1. Jacobs v. Salt Lake City School District, No. 23-4058 (10th Cir. Oct. 9, 2025). Full opinion via Justia.
2. FindLaw. Jacobs v. Council of Parent Attorneys and Advocates, Inc.; National Disability Rights Network; The Arc of the United States, Amici Curiae (2025).
3. Council of Parent Attorneys and Advocates (COPAA). (2026). Fact Sheet: The Least Restrictive Environment (LRE).
4. 34 C.F.R. §§ 300.114-300.117 (least restrictive environment requirements).
Every section above is a version of the same gap: good intentions, weak documentation.
IEP Builder, by The Lab, coaches SDI, goal quality, and compliance directly into the writing process.
See IEP Builder in action →
