Why We Built Parthion the Way We Did: A Whole-Child Early Warning System that Encompasses MTSS and Special Education
Two things happened this spring that, taken together, tell you everything about the moment we are launching into.
First, NPR reported on a Center for Democracy and Technology survey showing that 57 percent of special education teachers used AI to help develop individualized plans for their students during the 2024 to 2025 school year, up from 39 percent the year before. In a single year, AI went from the edge of the profession to the majority of it.
Second, New York City Public Schools, the largest district in the country, released its preliminary AI guidance. Its traffic-light framework, red for prohibited, yellow for proceed with caution, green for encouraged, put a hard line in the red zone: schools cannot use AI to develop IEPs for students with disabilities, cannot use it to make decisions about discipline, promotion, graduation, or placement, and cannot use it to surveil or counsel students. The guidance also states plainly that student data cannot be used to train AI models, that human review of AI output is always required, and that any tool touching student data must first clear a multi-step privacy and security review.
Read those two developments side by side and you see the central tension of education technology in 2026. Teachers are reaching for AI because they are drowning in compliance work. Regulators are drawing bright lines because the stakes for vulnerable students are too high to get wrong. Both groups are right. And most of the AI showing up in classrooms today resolves that tension in the worst possible way: an exhausted teacher pasting confidential student information into a consumer chatbot at six in the morning, with no privacy review, no guarantee the data will not be used to train a model, and no system of record behind it.
That is not transformation. That is a liability with a faster turnaround time.
We built Parthion for the other path, and we did not build it as outsiders. We built it because of who we are.
Parthion was built by special education parents and MTSS educators who believe that collaboration with families is the single most significant factor in better outcomes for ALL learners.
That sentence is our values statement, and it is also the design spec. So rather than open this launch with a feature list, I want to start with the four commitments we engineered into the product before we wrote a line of marketing copy. These are not afterthoughts. They are the architecture.
1. The humans are always the authors. Educators and families. Always.
Every guidance document worth reading says the same thing, and the NYC framework says it bluntly: human review of AI output is always required, and AI cannot replace the professional judgment of educators or the relationship between a teacher and a student. We agree so completely that we made it a structural property of the system rather than a policy we ask users to remember.
Parthion does not make decisions about children. In the IEP Sandbox, it generates evidence-cited recommendations for present levels, annual goals, and services, and every single line carries a confidence level and the cited evidence behind it. Then it stops. Educators accept, edit, or reject every recommendation before anything is committed to the IEP. The review queue is the product. And because our values put families first, the Sandbox is built to integrate family voice and teacher insight before and during the IEP meeting, collapsing weeks of back-and-forth into one collaborative, accountable workflow. The educator is the author. The team and the family own the document. The AI is a research assistant that never gets a vote.
This is also why I want to be direct about the IEP red line. Where a district’s guidance, like New York City’s today, places the development of the IEP itself off-limits to AI, Parthion is built to honor that, not route around it. The distinction that matters, and the distinction policy is moving toward, is between a tool that makes the decision and a system that surfaces a district’s own data, with cited evidence, so its people and the family at the table can decide. We built the second thing on purpose.
2. We never train models on student data. Ever.
This is the line in the NYC guidance that should reassure every parent in the city, and it is one of the first decisions we made as a company. To quote our own security standard: no student data is used to train external intelligence models, ever. A child’s disability status, evaluation results, and behavioral history exist to serve that child, not to improve a vendor’s product. When a district’s data enters Parthion, it stays the district’s data, encrypted at rest and in transit, and it never becomes training fuel.
3. Privacy is architecture, not advice.
The most quietly damning detail in the NPR story was the advice teachers give each other about consumer AI tools: do not enter anything that could identify a student. That advice only works if every teacher executes it perfectly, forever, under deadline pressure. It is a Band-Aid over a design flaw.
Our answer is not a better warning label. It is a platform built to the standards your district already requires: end-to-end encryption with industry-standard TLS and AES-256, granular role-based access scoped to district, school, grade, or caseload, and compliance with FERPA, CIPA, COPPA, SOPIPA, and New York State Education Law Section 2-d, with SOC 2 Type II underway. It is designed to clear the multi-step data privacy and security review that NYC and other districts now require before any tool touches student data. We would rather a district’s review process be hard to pass. That is the point.
4. No surveillance. No autonomous calls on kids. And a higher bar we set ourselves.
The NYC framework prohibits using AI to surveil or counsel students, and to make consequential decisions about discipline or placement. Parthion does none of those things, by design.
But I want to name one more thing, because it is where responsible vendors should lead rather than wait. NYC’s officials conceded that their current review process does not yet evaluate AI tools for algorithmic bias or for effectiveness in the classroom. In special education, where Black and brown students have been disproportionately identified and misplaced for decades, an unaudited algorithm is a risk multiplier. So we built Parthion’s intervention library on NCII-vetted practices with documented effect sizes, aligned every recommendation to ESSA evidence tiers, and cite What Works Clearinghouse findings alongside each suggestion so educators see the evidence basis before they decide. If we are going to ask districts to trust this technology with their most vulnerable students, we should be the ones holding the higher bar.
Two modules. One platform. Identify sooner, act faster.
Hold those four commitments in mind, and the question becomes practical: what does a compliant, human-verified, whole-child system actually do for a district?
Parthion unifies MTSS and special education in one platform. On the MTSS side, it applies your district’s own risk thresholds, refreshed nightly, to whole-child data across academics, attendance, behavior, and wellness, and assigns students to the right tier at the right time. It builds structured intervention plans with dosage, frequency, and fidelity tracking, and it runs OLS-regression decision analysis against the goal line so a team knows whether an intervention is on pace, off track, or losing ground. It even flags a fidelity gap before you intensify, because if delivery dropped to 50 percent, the missed sessions may be the real problem, not the plan. The result is accountability most districts have never had: leaders can prove supports were actually delivered, not just assigned, so that “assigned” and “delivered” finally stop being the same word.
On the special education side, the IEP Sandbox carries that same data forward, from a referral workflow with Child Find flags and the New Jersey 20-day clock, to evidence-cited drafts of PLAAFPs, goals, and services, to FAPE- and LRE-aligned, signature-ready exports. And it all connects, because this is the argument I have made my entire career. An IEP is a Tier 3 document, the most intensive end of a multi-tiered system of support. It should never be built in a vacuum. When the same system that flags risk, matches interventions, and monitors fidelity is the one informing the team’s work at Tier 3, the plan finally reflects the child, because the system can finally see the child.
The bottom line
The 57 percent number is not a problem to manage. It is a demand to be met responsibly. Our special educators have told us, with their own behavior, that the compliance burden has become unsustainable and that AI can relieve it. The regulators have told us, with their red lines, exactly where the guardrails belong. Parthion exists in the space where those two truths agree: a Whole-Child Early Warning System that gives educators their time back without ever giving up human judgment, student privacy, or the trust of the families we serve.
We have spent years in this field saying we want to see every student. We built Parthion to finally do it, the right way, from the first line of code. It is the missing layer between data and action.
Ready to close the gap? Schedule a 20-minute walk-through at parthion.io
Nick Gronda is the Founder and CEO of Parthion. He previously served as CEO of College ROI.