Learning and Schools
Schools are where most people first learn to think alongside other people. This chapter covers K-12 classrooms and teachers, special education and the team that writes an Individualized Education Program (IEP), small and rural districts, tutoring and peer learning, adult literacy, community colleges and apprenticeships, libraries, and the family's relationship to the school. The pull toward a personal AI tutor for every child, alone at their own pace, is strong and well funded. The chapter lays out the working alternative, where AI takes the grading and the paperwork so teachers lead small groups and students explain ideas to each other, then maps nine territories with facts, plays, and scores.
Index: Worth Building
Teachers lead, students explain, the paperwork disappears
Learning has always happened among people. A student explains a proof to a friend and understands it for the first time. A teacher sees who is struggling because she is moving between tables rather than grading. A parent, a teacher, and a therapist decide at one table what a child needs this year. Each of these is being replaced by a screen pointed at one person. None requires less AI to protect. They require AI pointed at the paperwork, the drills, and the scheduling, so the people have time for each other.
| Scene | The default path | What we could build instead |
|---|---|---|
| The classroom | Every child has a personal AI tutor and learns alone at their own pace. The teacher watches dashboards. | AI grades and runs the drills overnight. The teacher spends the period at small tables. Students explain the hard idea to each other, with prompts the AI prepared. |
| The IEP meeting | A forty-page plan arrives in English. A chatbot summarizes it. Services slip when a position goes unfilled, and nobody records the gap. | The team shares one plain record in the family's language, with hours owed and delivered visible every month. The family can export the plan and take it anywhere. |
| The small district | A district of 400 students buys the same vendor stack as a city, pays for parts it cannot use, and cannot get its data back. | Districts own a shared back office through their regional service agency. The schema is open, the rules are versioned by state, and leaving costs nothing. |
| The absent student | A robocall on day three. A truancy letter on day ten. A court referral on day eighteen. The pattern was visible on day two. | A small circle the family agreed to sees the pattern early. A person knocks. The district sees absence by school and month and acts on the trend. |
| The adult learner | A literacy app on a phone, alone, with streaks and badges. Enrollment in programs with people keeps falling. | A volunteer tutor and a learner meet at the library. The AI built the lesson from the learner's own lease and school letters. The pair does the reading. |
| The career record | A platform owns the skills profile. The learner sees it only while subscribed, and the employer sees what the platform decides. | The learner holds a record that the apprenticeship sponsor, the community college, and the employer all read in the same format, and can leave with it. |
What each commitment means when the subject is a school
| Commitment | In learning and schools, it means | A concrete instance |
|---|---|---|
| 1 · AI is a means, not the measure | Judge a school tool by what students can explain to each other and what teachers do with the returned hours, not by minutes on task. | Teachers who use AI weekly reported saving 5.9 hours a week in Gallup's April 2025 survey. The tool should also report time spent in small groups. |
| 2 · Keep people in charge | The teacher decides who sits together and what gets retaught. The IEP team decides services. Families can see the plan, change it, and export it. | Ed-Fi Data Standard v6.1, released 7 May 2026, models the IEP as a first-class, time-bound document with amendments and history. An export in that format is the exit door for a family. |
| 3 · Build capability, not dependence | Students who used a tool must do as well or better without it. Tutors who used a copilot must become better tutors. | Tutors using Stanford's Tutor CoPilot were 10 percentage points more likely to ask students to explain their thinking. |
| 4 · Make drift visible | Surface the slow losses. Service hours not delivered, absence creeping up, adult education enrollment falling for twenty years. | One Connecticut complaint found a student missed 308.75 hours of special education in 2025-26 because no certified teacher was assigned. A monthly owed-versus-delivered line would have shown it in October. |
| 5 · Design for groups | The class, the study circle, the IEP team, the tutoring pair, the family and the school, the cooperative of small districts are the users. | A trial of about 4,000 students put children in groups of four, alternating between a human tutor and software every other day, and gained 0.23 standard deviations in math. |
| 6 · Protect the friction that matters | Remove the grading, the compliance forms, the scheduling, the translation queue. Keep the discussion, the explaining, the meeting, the knock on the door. | California requires translated IEPs within 30 calendar days for a district's eight most common languages. Automate and check the translation. Keep the meeting human. |
| 7 · Refuse inevitability | The solo tutor, the grade-alert app, and the vendor-locked record are choices. Peer tutoring, service agencies, and public libraries are alternatives with decades behind them. | Nine meta-analyses put peer tutoring's effect between 0.33 and 1.05 standard deviations. 553 regional service agencies share services in 44 states. |
The three questions, applied
| Thing | More capable or more dependent | Closer to each other or to the system | See, steer, and leave |
|---|---|---|---|
| A personal AI tutor used alone at home | Dependent in the trials that tested it without a teacher. Short-term gains, worse unaided. | Alone with a screen | Can usually leave; the learning record stays with the vendor |
| A human tutor with a copilot | Capable. Weaker tutors improved most, and students explained more. | Closer. The pair is the unit. | The tutor sees every suggestion and can ignore it |
| A parent-only IEP navigator subscription | Capable. The parent learns the rules. | Closer to an advisor; the school is on the other side | Can cancel; the record is the parent's |
| An IEP record shared by the whole team, with an hours ledger | Capable. Everyone can explain the plan. | Closer. The team is the user. | Export everything in an open standard |
| AI cheating detectors | Dependent on a verdict nobody can audit | Teacher against student | Opaque; half of students already fear false accusation |
The five places to push, in a school system
| Place to push | What it means | In learning and schools |
|---|---|---|
| Paradigm | AI as a means | Learning as something people do together, with explaining as the core act. The tool buys time for discussion and carries the paperwork. |
| Goals and metrics | Capability, trust, connection | Minutes of student talk per period; students who can explain the idea without the tool; service hours delivered as a share of owed; attendance by month; adult learners who stay past week twelve. |
| Information flows | Make drift visible; the most practical lever | Owed versus delivered special education hours; absence by school and month; where returned teacher hours go; dispute rates by state; adult education enrollment over time. Most of it exists in fragments and reaches nobody who can act. |
| Rules | Legibility, portability, interoperability, exit | IEP export in an open standard; a plain register of what every classroom AI does; family message histories that export; learner-owned records; district AI policies as versioned, readable rules. |
| Self-organization | Who can act together | Service agencies and purchasing cooperatives for small districts; parent groups and the IEP team; tutoring providers sharing a copilot commons; library study circles; sponsors and colleges on a shared record. |
The orientation document says cultural orientations change when people see an alternative work, and offers Wikipedia and cooperatives as proof. The one-room school had older children teach younger ones. The public library was built as a shared room for self-education. The parent-teacher association is a century-old civic body in nearly every town. Regional service agencies are cooperatives in all but name, created so small districts could share what none could afford alone. The task is to point the new tools at the paperwork and the drills, and to protect the explaining, the meeting, and the knock on the door.
Nine territories in learning and schools
Scored on the shared rubric of shared layer, buyer or deadline, outcome data, the three questions, and fit. Each card names the reusable parts it draws on. Figures from advocates or vendors are labeled.
How to read a card: its score adds five judgments, each from 0 to 2 (shared layer nobody owns, buyer or deadline, outcome data, the three questions, fit), for a total out of 10.
The IEP as a shared, portable record, with hours owed and hours delivered
The 7.5 million students served under the Individuals with Disabilities Education Act (IDEA), their families, and the professionals on their teams. The team is the user.
Why now
Students served under IDEA reached 7.5 million in 2022-23, 15 percent of public school students. Due process complaints rose to 40.1 per 10,000 children served in 2021-22, up from 32.7, and the Center for Appropriate Dispute Resolution in Special Education's (CADRE's) database allows state comparisons back to 2004. In 2024-25, 45 states reported special education as a shortage area. Connecticut's 2025-26 complaint summaries include a student owed 239.4 hours who received 12.5, and another who missed 308.75 hours while no certified teacher was assigned. On 7 May 2026, Ed-Fi Data Standard v6.1 introduced an early-access Special Education model with a StudentIEP entity and an IDEAEvent entity for referrals, evaluations, consent, and meetings. Parent-side navigators treat the parent alone as the user. Undivided serves California at $95 to $115 a month for its navigator tier, and Wayfound covers 71 programs in California and Texas.
Plays
- Export my IEP. A family-held copy of the plan, its amendments, and its history in the Ed-Fi StudentIEP format, readable by the next district, the pediatrician, and the family itself.
- An hours ledger the whole team sees. Services owed against delivered, by month, in the family's language, with the compensatory calculation already done when a position goes unfilled.
- A meeting kit for the team rather than a brief for one side. Who will be there, what the data says, what the parent wants raised, the deadlines, translated and checked by a person.
Already there
Undivided and Wayfound on the parent side. The Ed-Fi Alliance on the schema. CADRE on dispute data. Connecticut publishes complaint summaries with hours owed and delivered. Nobody holds the team-side record or the ledger.
Score
Shared layer 2 · Buyer or deadline 2 · Outcome data 2 · Three questions 2 · Fit 2
Grading and drills off the desk, discussion onto it
The class as the user. Teachers who work 54 hours a week, and students who learn most when they explain the idea to a classmate.
Why now
A 2025 literature review by TNTP puts the typical teacher's week at 54 hours, 43 at school and 11 at home, with instruction taking at least 40 percent and planning about 20 percent. In April 2025, Gallup found 6 in 10 public school teachers had used AI tools that year, and weekly users reported saving 5.9 hours a week. Only 32 percent used AI weekly, and only 18 percent said they had received formal guidance from administrators on using AI at work. The students who do the tutoring in peer programs gain as well.
Plays
- A time-returned ledger. The tool that saves the hours also records where they went, in categories teachers choose, and the school publishes the totals. Small groups and family calls count; more screen time does not.
- A protocol library for explaining. Open, teacher-edited routines for pairs and small groups, with AI preparing the prompts, the pairings, and the misconception list.
- A grading assistant that shows the teacher the pattern across the class so she decides what to reteach and who sits together. The student never meets the model alone.
Already there
Commercial teacher assistants for lesson prep and grading. Stanford's National Student Support Accelerator on explaining prompts. No ledger of returned time, and no open protocol library.
Score
Shared layer 1 · Buyer or deadline 2 · Outcome data 2 · Three questions 2 · Fit 2
Tutoring that makes tutors better, after the federal money ended
The tutor and the small group as the user. Tutoring providers, students who lost programs when relief ended, and districts deciding what to keep.
Why now
Stanford's Tutor CoPilot trial embedded an AI assistant with 900 tutors serving 1,800 students. Mastery rose 4 percentage points overall and 9 points for students with lower-rated tutors. The counter-evidence is just as clear. Turkish students given plain GPT-4 solved more practice problems and then scored 0.19 standard deviations below control without it, while a tutor-prompted version scored 0.01 below. In Nigeria, a six-week program with teacher support gained about 0.3 standard deviations. In March 2025 the Department of Education rescinded extensions for more than $2.5 billion of unspent relief, and Baltimore ended tutoring and after-school programs serving about 4,000 students.
Plays
- A copilot commons. Open prompts, misconception libraries, and evidence that community tutoring programs can adopt without a vendor, measured by explain-prompts per session and tutor growth.
- A shared outcome registry for tutoring providers by district and month, with consent, so a board deciding what to keep can see dosage and results side by side.
- Cross-age peer tutoring at scale. Scheduling, pairing, and prep handled by the system; the tutoring done by the older student, who learns by teaching.
Already there
The National Student Support Accelerator, Saga Education, and the Tutor CoPilot team. J-PAL is running a coaching trial in New Mexico. The commons and the registry are open.
Score
Shared layer 1 · Buyer or deadline 1 · Outcome data 2 · Three questions 2 · Fit 2
Absence made visible early, with a circle around the student
The roughly 10.8 million students chronically absent in 2024-25, their families, and the counselors who meet them at day eighteen instead of day two.
Why now
In 2024-25, 22.6 percent of students in 44 states and the District of Columbia were chronically absent, according to an American Enterprise Institute analysis reported in June 2026. That is down 0.9 points from 23.5 percent in 2023-24, and from a high of 28.3 percent in 2022, but still far above the pre-pandemic rate of about 15 percent. RAND's August 2025 estimate was about 22 percent, or 10.8 million students. The default response is robocalls, letters, and, in some states, court.
Plays
- A consented circle around the student. Family, a counselor, a coach, a relative. The system spots the pattern in week one and a person reaches out. No letter goes out unread by a person.
- A notice reader for attendance and discipline letters. What this letter means, which deadline it carries, which right the family has, in the family's language.
- An absence atlas by school and month, built with districts, so the trend is public before it is normal.
Already there
Attendance analytics vendors. RAND and the American Enterprise Institute (AEI) on the national picture. The circle around the student does not exist as a product.
Score
Shared layer 1 · Buyer or deadline 2 · Outcome data 2 · Three questions 1 · Fit 1
The small and rural district's shared back office
The 7.3 million students in rural districts, most of them in districts under 493 students, and the service agencies those districts already belong to.
Why now
Why Rural Matters 2023 counted 7.3 million students in rural districts in 2022-23, about 14 percent of all students, and 64 percent of rural districts enroll fewer than the median of 493. 553 educational service agencies operate in 44 states with combined budgets of about $15 billion. In Nebraska's 2025-26 vacancy survey, special education remained the highest need at 140 unfilled positions, and 60 percent of districts named no applicants as their main problem. The Community Eligibility Provision fed 27.2 million children in 2024-25, and 74.9 percent of eligible schools adopted it for 2025-26. Eligibility rests on data matching to the Supplemental Nutrition Assistance Program (SNAP) and Medicaid, so paperwork capacity decides whether a small district qualifies.
Plays
- A member-owned data and compliance office under the service agency. State reporting, federal paperwork, and meal certification drafted by models and decided by the district, in a schema the districts own.
- Shared special education staffing across a region, so a district of 300 students gets a speech therapist on Tuesdays and the L1 ledger shows it.
- A standards cohort of small-district software tools adopting Ed-Fi and shared formats, then winning a state buyer.
Already there
The Association of Educational Service Agencies and its members. The Food Research and Action Center on community eligibility. No cooperative data office exists at the scale of a service agency.
Score
Shared layer 2 · Buyer or deadline 1 · Outcome data 1 · Three questions 2 · Fit 2
The family and the school, in the family's language
The family and the school as a pair. Parents who receive letters they cannot read, alerts they did not ask for, and a weekly flood of numbers about their child.
Why now
TalkingPoints reports 24 million educators and families reached since 2015, a vendor figure. ParentSquare reports serving 20 million students in all 50 states with translation in more than 190 languages, also a vendor figure. California asked its Department of Education for IEP translation guidance by 1 July 2025. By mid-2025, 22 states and the District of Columbia had enacted laws or policies restricting student phone use.
Plays
- A notice reader for school mail. A letter or PDF in; who it is about, what it asks, which dates, and where to read the rule come out, in the family's language.
- A weekly plain-language digest that replaces hourly alerts, with one suggested thing to talk about at dinner. The family chooses the cadence and can turn alerts off without losing the channel.
- Portable conversation history. The family's messages with the school export when the family or the district changes vendors.
Already there
TalkingPoints and ParentSquare on messaging and translation. Nobody builds the digest, the export, or the reader for letters that still arrive on paper.
Score
Shared layer 1 · Buyer or deadline 1 · Outcome data 1 · Three questions 2 · Fit 2
Adult literacy circles at the library
The learner and the volunteer tutor as a pair, and the public library as the room they meet in. More than one in four American adults reads at the lowest levels.
Why now
In the 2023 cycle of the Program for the International Assessment of Adult Competencies (PIAAC), released in December 2024, 27 percent of U.S. adults scored at Level 1 or below in literacy, up from 18 percent in 2012. Federally funded adult education served more than 2.6 million students in 2000 and fewer than 900,000 by 2021. In the Public Library Association's 2023 technology survey, published in 2024, 95.3 percent of public libraries offered digital literacy training, 46.9 percent lent internet hotspots, and almost 30 percent ran a digital navigator program.
Plays
- AI prepares the tutor, the pair does the reading. Lessons built from the learner's own documents, a lease, a school letter, a pay stub, with the tutor seeing every suggestion.
- Study circles scheduled by the library. The system handles matching, rooms, reminders, and progress notes. The meeting is in person.
- A consented outcome commons for adult programs by county and month, so twenty years of falling enrollment becomes visible to funders.
Already there
ProLiteracy and its member programs. The Public Library Association and digital navigator programs. The tutor-preparation layer and the commons are open.
Score
Shared layer 1 · Buyer or deadline 1 · Outcome data 2 · Three questions 2 · Fit 1
The learner-owned record across apprenticeship, college, and work
The 703,475 registered apprentices active at the end of fiscal 2025, community college students trying to transfer, and the sponsors, colleges, and employers who each keep a separate record of the same person.
Why now
At the end of fiscal 2025 there were 703,475 active registered apprentices, 311,035 new apprentices, 119,035 completions, against a federal goal of one million active apprentices. Community college enrollment grew 3.0 percent in fall 2025, and undergraduate certificate enrollments at community colleges reached 752,000, up 28.3 percent from fall 2021. Transfer enrollment grew 4.4 percent in fall 2024, yet two-year to four-year transfer rates have not recovered to pre-pandemic levels. Alabama's Talent Triad reports nearly 132,000 workers demonstrating skills through Learning and Employment Records in 14 months, a state program figure. California's May 2026 scan for a Career Passport counted seven existing wallet programs across five states, Pittsburgh, and Arizona State University.
Plays
- An open export that an apprenticeship sponsor, a community college registrar, and an employer all read, held by the learner, so leaving a platform or a state does not mean starting over.
- A transfer evidence file. What the student completed, what the receiving college accepts, and the appeal when credit is denied.
- A standards cohort of small training providers and sponsors adopting shared credential formats, then winning a state buyer among the programs already building wallets.
Already there
Alabama Talent Triad and its credential registry. California's Career Passport effort. The learner-held export is open.
Score
Shared layer 2 · Buyer or deadline 2 · Outcome data 1 · Three questions 2 · Fit 2
School AI policy as a public, legible, versioned choice
Students, teachers, and principals in districts where most people now use AI for school and most have no written rules. The district and its community are the user.
Why now
RAND's 2026 report found more than half of students and teachers using AI for school, with use up about 15 percentage points among students and 25 among teachers in a year. 26 states offered guidance, under half of principals said their school or district had a policy, and about a third of teachers had a policy specific to academic integrity. Gallup found schools with a policy saw a 26 percent larger time dividend, 2.3 hours a week against 1.7. Half of students feared false accusation. The rules are being written this school year, district by district, mostly in private.
Plays
- Open, versioned policy configuration per district, in plain language, that a family can read in five minutes and a neighboring district can copy.
- A legibility register. For every AI tool in the building, what it does, what it sees, who can override it, and how a student appeals. Published on the district site.
- Choices named as choices. A short public comparison of the solo-tutor path and the small-group path for the board meeting where the purchase is decided.
Already there
State guidance in 26 states. RAND and Gallup on the gap. No district publishes a register, and no shared policy format exists.
Score
Shared layer 2 · Buyer or deadline 2 · Outcome data 1 · Three questions 1 · Fit 2
The orientation document's education row is the clearest in its table. On one side, every child alone with a tutor. On the other, teachers in small groups and students explaining to each other. Plain chatbot access made Turkish students worse without it. A copilot made weak tutors better. The pattern worth building is the one where the model prepares the adult and the group, carries the paperwork, and never becomes the only one the child talks to. Offer it to teachers, tutoring programs, IEP teams, and libraries rather than to children.
A Tuesday in a district of nine hundred students
At 7:40 a teacher opens her laptop and finds last night's exit tickets already sorted. Fourteen students made the same fractions mistake. The groups for second period are proposed, mixed so each table has one student who got it. She changes two names. During the period she sits at four tables in turn while a girl explains common denominators to the boy beside her, badly at first, then clearly. The teacher hears the moment it clicks because she is there.
At 10:15 the IEP team meets. The mother has the plan on her phone in Spanish, with this fall's speech sessions owed and delivered side by side. Three were missed in September when the therapist was out. The compensatory time is already proposed. They spend the hour on what the boy should be able to do by May.
At noon the business manager approves the state report the service agency's system drafted overnight, and the meal data match that keeps lunch free next year. At 4:00 a retired nurse meets her adult learner at the library. The lesson is built from his daughter's school letter, and by 4:40 he has read it to her. At 7:00 a school board reads a one-page register of every AI tool in the building before it votes. Nobody in this day used less AI than the default path. The AI was pointed at the paperwork.
Signs we are getting there
- Teachers who use AI tools report where the saved hours went, and the share spent in small groups rises year over year in a published ledger.
- A family can download a complete IEP, with amendments and history, in the Ed-Fi StudentIEP format from at least one state's districts.
- At least one state publishes special education hours owed against hours delivered by district and month, and the gap shrinks.
- Chronic absenteeism falls below 20 percent nationally, and the districts that got there used a person-first outreach model that others can copy.
- Students in classrooms using AI score as well or better on unaided tests than students without it, in independent trials.
A weekend, a season, a career
A weekend
- Pull a year of your state's special education complaint summaries, as Connecticut posts them, and tally hours owed against delivered. Publish the table. (L1)
- Ask your district for its AI policy and the list of AI tools in use. Write a one-page plain-language version and send it to the board. (L9)
- Spend a Saturday as a library digital navigator or literacy tutor. Log what the learner brought in to read. (L7)
- Find the service agency that serves your county and document what it offers small districts and what it charges. (L5)
A season
- With one school, run a 90-day time-returned ledger in which teachers log where AI-saved hours went, in categories they choose. The output is a public table of hours by use. (L2)
- With a parent group and one district, export ten IEPs into the Ed-Fi StudentIEP format and build their owed-versus-delivered ledger. The output is ten families who can read and leave with their plan. (L1)
- With a tutoring provider, pilot an open copilot prompt library across twenty tutors. The output is explain-prompts per session, before and after, and tutor ratings. (L3)
- With a library system, run a twelve-week literacy circle using AI-prepared lessons from learners' own documents. The output is retention past week twelve and learner-named goals met. (L7)
A career
- Lead the standards cohort for IEP portability, funding small special education tools to adopt the Ed-Fi model, then winning a state as the buyer. (L1)
- Found a member-owned data and compliance cooperative under a regional service agency, with small districts as the owners. (L5)
- Convene the open learner-held record that apprenticeship sponsors, community colleges, and employers all read, starting with the programs already building wallets. (L8)
- Run a district's AI legibility program, publish the first register, and make the format one other districts copy. (L9)
Who is already doing it
- National Student Support Accelerator, Stanford. Ran the Tutor CoPilot trial and tracks tutoring after federal relief ended.
- Saga Education. Group math tutoring, now in a trial pairing tutors with AI-supported coaching.
- Undivided. A parent-side navigator for children with disabilities in California, with a paid navigator tier.
- Ed-Fi Alliance. Maintains the open K-12 data standard that now models the IEP as a first-class record.
- CADRE. Publishes IDEA dispute resolution data by state back to 2004.
- Association of Educational Service Agencies. The national body for 553 regional agencies serving small districts.
- TalkingPoints. Two-way translated messaging between teachers and families, used since 2015.
- Alabama Talent Triad. A state skills wallet, credential registry, and job tool used by about 132,000 workers.
Nine territories, scored
Score key: SL shared layer nobody owns, BD buyer or deadline, OD outcome data accrues, 3Q the three questions, FT fit for a builder who convenes. Each is scored 0 to 2, so the total is out of 10.
| Code | Territory | Shared layer | Buyer or deadline | Outcome data | Three questions | Fit | Total | Horizon |
|---|---|---|---|---|---|---|---|---|
| L1 | The IEP as a shared, portable record | 2 | 2 | 2 | 2 | 2 | 10 | Now, with one district and a parent group |
| L2 | Grading and drills off the desk | 1 | 2 | 2 | 2 | 2 | 9 | 2026-27 school year, one school |
| L8 | The learner-owned record | 2 | 2 | 1 | 2 | 2 | 9 | 2027 cohort for a state buyer |
| L3 | Tutoring that makes tutors better | 1 | 1 | 2 | 2 | 2 | 8 | 2027 with a provider |
| L5 | The small district's shared back office | 2 | 1 | 1 | 2 | 2 | 8 | 2027 with a service agency |
| L9 | School AI policy as a legible choice | 2 | 2 | 1 | 1 | 2 | 8 | Now, this policy season |
| L4 | Absence made visible early | 1 | 2 | 2 | 1 | 1 | 7 | 2027 with one district |
| L6 | The family and the school | 1 | 1 | 1 | 2 | 2 | 7 | 2027 |
| L7 | Adult literacy circles at the library | 1 | 1 | 2 | 2 | 1 | 7 | 2027 with a library system |
Six to pursue first, and five to decline
- The IEP as a shared, portable record (L1). A new open standard dated May 2026, a rising dispute rate, published evidence of hundreds of missed hours per child, and no shared tool for the team. Start with ten families and one district this year.
- Grading and drills off the desk (L2). The orientation document's own example, now with survey evidence that hours are being saved. Build the ledger that shows where they go, and the protocol library.
- The learner-owned record (L8). Seven wallet programs already exist and a federal apprenticeship goal creates pull. Convene the export format before each state locks its own.
- Tutoring that makes tutors better (L3). The trials point one way. A copilot commons and an outcome registry help the programs that survived the end of relief money.
- The small district's shared back office (L5). The cooperative already exists in 553 places. Add a member-owned data office and the meal data match, and leaving a vendor becomes possible.
- School AI policy as a legible choice (L9). Most districts are writing rules this year with no shared format. A readable register and a versioned policy template are cheap and copy fast.
- Tutoring and AI evidence: Stanford NSSA, Tutor CoPilot study; technology-assisted high-dosage tutoring trial; Bastani and colleagues, generative AI can harm learning; World Bank on the Nigeria program; J-PAL, Saga coaching trial in New Mexico; NSSA on tutoring after relief funds ended
- Peer tutoring: review of nine peer tutoring meta-analyses
- Teacher time and AI use: Gallup and Walton Family Foundation, April 2025 survey; TNTP, teacher time use literature review, 2025
- Teacher and special education shortages: The Progressive, May 2026, on special education shortages in 45 states; Nebraska 2025-26 teacher vacancy survey
- District AI policy: coverage of RAND's 2026 report on AI use and policy gaps
- Chronic absenteeism: Chalkbeat on the AEI analysis, June 2026; RAND, chronic absenteeism in 2024-25
- Special education data and disputes: Ed-Fi Special Education data model overview; Ed-Fi, what is new in v6.1; CADRE national dispute resolution data summary; National Center for Education Statistics, students with disabilities, 2022-23; Connecticut complaint 26-0697, New Haven; Connecticut complaint 26-0550, East Hartford; California SB 445 analysis on IEP translation
- Parent-side navigators: Undivided pricing; Wayfound
- Small and rural districts: AESA, 553 agencies in 44 states; Why Rural Matters 2023; FRAC on community eligibility
- Adult literacy and libraries: Institute of Education Sciences release on the 2023 PIAAC results; LEO summary of U.S. PIAAC 2023 results; ProLiteracy on declining adult education enrollment; American Library Association, 2023 Public Library Technology Survey
- Community colleges, apprenticeship, and records: Clearinghouse, preliminary fall 2025 enrollment; Clearinghouse transfer report, fall 2024; Michigan labor market news, FY2025 apprenticeship totals; Work Shift on the one million apprentices goal; Work Shift on Alabama Talent Triad; California Career Passport executive summary, May 2026
- Families, phones, microschools, relief funds: TalkingPoints 2025 reflection (vendor); ParentSquare figures (vendor); Rockefeller Institute on state phone laws, 2025; The 74 on microschools in 2026; EdChoice 2025 participation ranking; Center for American Progress on the March 2025 rescission
Assumptions and limits. Scores are judgments. Figures from TalkingPoints, ParentSquare, the National Microschooling Center, EdChoice, and Alabama's Talent Triad come from vendors, advocates, or the program itself and are labeled as such. The Ed-Fi, CADRE, Food Research & Action Center (FRAC), Undivided, and Wayfound facts were verified by the chapter lead and re-cited here; direct fetches of those pages were blocked in this pass. The PIAAC share at Level 1 or below appears as 27 percent in the summaries used here; National Center for Education Statistics (NCES) tables may round differently. A widely circulated one-in-eight teacher shortage figure was left out because its original source could not be confirmed. The 54-hour teacher week is a literature-review figure from surveys before 2025. The phone-law count is as of mid-2025 and has since grown. The Turkish trial's student count was omitted because sources disagreed. Absenteeism rates differ by source and state coverage, so AEI and RAND figures are both given with dates.