OKO is an AI-powered small-group math platform for grades 3–12. Below you’ll find the privacy, data, accessibility, and evidence answers you need, each with documentation to back it up.
Requirement
Where OKO Stands
Document
FERPA and COPPA
Compliant. Meets or exceeds student-privacy standards in all 50 states.
State and district data review
Cleared the NYC Department of Education's ERMA review.
Available on request
Student data and AI models
De-identified before reaching any third-party model. Providers are contractually barred from training on it.
Encryption
Encrypted in transit and at rest.
Accessibility
WCAG 2.2 Level AA target. Accessibility Conformance Report based on VPAT 2.5 Rev, August 2026.
Evidence
ESSA Tier III and IV (LeanLab certificates, WestEd feasibility study). Independent WestEd randomized controlled trial.
Rostering and sign-in
No rostering required. Students join by group code. Teachers sign in with Google, Clever, ClassLink or Microsoft.
WestEd ran a randomized controlled trial of OKO in 11 schools in Hamilton County Schools, Tennessee: 20 fourth- and fifth-grade teachers and 1,040 students, randomly assigned by classroom, January to April 2026. WestEd designed the study and ran every analysis independently.
~1.5
months of added math growth
Students who used OKO gained approximately 1.5 months of additional growth on MAP over ten weeks, compared with comparison students.
Exploratory: teachers chose which students OKO served after randomization, so WestEd calls this a strong signal rather than proof. +2.46 RIT above NWEA projection, p = 0.033, g = 0.31.
2.8
hours back each week
Teachers reported OKO gave them back 2.8 hours a week, against 1.0 hour for the software comparison teachers already used.
Teacher-reported on a post-survey (10 treatment, 8 comparison teachers), not a measured time study.
WestEd's observers documented students in OKO classrooms explaining their reasoning and correcting each other's errors. In comparison classrooms, students were on task but rarely talked about math with one another.
Hamilton County teacher, via WestEd
OKO holds ESSA Tier III and IV certificates issued by LeanLab Education on the WestEd feasibility study: a peer-reviewed study of about 115 students that found significant pre-to-post MAP gains (p = .006) and reduced math anxiety.
The Hamilton County randomized trial is a separate study, and we attach no ESSA tier to it.
Students in an OKO session talk to each other, and OKO listens so it can facilitate. So we're precise about what gets captured and what happens to it.
Transcripts only. No raw audio or video is kept.
De-identified before any model. Student data is stripped of identifiers before it reaches a third-party AI model, and those providers are contractually prohibited from training on it.
Encrypted in transit and at rest.
FERPA and COPPA compliant, meeting or exceeding student-privacy requirements in all fifty states.
Cleared NYC DOE's ERMA review, one of the highest bars in the country. We'll share the documentation before any pilot conversation.
Most AI in school puts one student alone with a chatbot. OKO does the opposite. It facilitates the conversation between students in a group of two to five.
There is no one-to-one conversation between a student and a model anywhere in the product. OKO nudges the group: bringing in a quieter voice, asking someone racing ahead to explain, surfacing a hint only when the group is stuck. It doesn't hand over answers.
The math happens on paper and whiteboards. The screen carries the problem, the group's shared work and the facilitation.
The teacher is there the whole time, circulating, with a live view of every group. OKO runs the groups the teacher isn't sitting with.
If your district has published screen-time guidance, OKO's dosage sits inside most of it. The U.S. Department of Education's August 2026 Dear Colleague letter asks districts to judge classroom technology by instructional value and learning outcomes rather than minutes used. We'd ask you to hold OKO to that standard.
minutes per student per week
Two or three sessions of 15 to 20 minutes. Below that window the academic benefit softens; above it, discourse quality drops. We track weekly usage in every partner classroom and report it back to school leaders.
grade floor
OKO isn't recommended below grade 3. The youngest learners shouldn't be doing collaborative work on screens.
A product premised on every student having a voice in the conversation has to be usable by every student.
OKO targets WCAG 2.2 Level AA, which incorporates the WCAG 2.1 Level AA requirements referenced in state K–12 accessibility statutes such as Colorado HB21-1110. Every student-facing and teacher-facing workflow can be completed with a keyboard alone and with a screen reader. Problem text and facilitator prompts are read aloud through built-in text-to-speech, and students participate by speaking as well as typing.
To report a barrier, email support@okolabs.ai with "Accessibility" in the subject line. We aim to acknowledge within three business days.
VPAT 2.5 Rev, August 2026. Based on an evaluation conducted June 15–17, 2026 with Axe DevTools, Lighthouse, keyboard-only navigation, and ChromeVox and VoiceOver screen readers.
A self-evaluation. Where the evidence supported two readings, we recorded the lower level. An independent third-party audit is planned.
Six criteria reported as Partially Supports, each with a remediation target. Three concern third-party problem content under active audit; the others are reflow at 400% zoom, one footer contrast element, and live captioning of student speech.
Kept current. Automated accessibility checks run on every release, manual screen reader passes run quarterly, and the report is reissued at least annually.
Tell us what you need and we'll send OKO's privacy, security and accessibility documentation, or walk your team through any of it on a call.
