Skip to content
coeducate
Sign in
How it works Features Use cases Pricing Student privacy Blog FAQ Contact Sign in

For principals, coaches and district leads

AI pilot programs for schools that end with per-student evidence, not a teacher survey

Run an eight-week AI tutoring pilot in two to four classrooms, at the normal published price, and walk into the decision meeting with a dated record of what each student got wrong and whether it changed.

See pricing

Two teachers for two months is $356. Student work is never used to train models.

Mastery Check

of 3 runs left

Subject

Grade band

Nothing here is stored

Use the sample or your own practice text. Never paste a real student's name.

Tutor mode

Session evidence card

This is what a teacher opens after every session. That was your third run. Nothing you typed was stored.

The short answer

A pilot is only useful if you decide the pass mark first

An AI pilot program in a school is a time-boxed trial of one tool in a few classrooms, with a buying decision at the end. The pilots that produce a clear yes or no pick one problem, agree on the result that would justify the spend before week one, and collect that result from student work. The ones that drift end with a survey.

Coeducate is an AI tutor for grades 3 to 12. A student works a problem from your own lesson, the tutor coaches with hints instead of answers, and every session ends with an evidence card naming the skill, quoting the student and suggesting a reteach. That card is the pilot data. Nobody has to build a spreadsheet for it.

01 What the public pilots reported

Four state AI pilots and what they could show at the end

The Education Commission of the States summarized the K-12 AI pilots states had run or funded by mid-2025. The pattern is worth noticing before you plan your own. The largest one, in Indiana, reached 112 schools in 36 districts, and its headline result was a teacher opinion survey.

Indiana

AI-Powered Platform Pilot Grant, 2023-24

One-year grants paid subscriptions and PD for AI tutoring and workload tools, about $2 million in federal relief funds.

A teacher survey: 53% rated the impact on students and teaching positive or very positive.

Connecticut

AI Pilot Program, spring 2025

Seven districts, grades 7 to 12, state-approved AI tools with teacher PD, required by Public Act 24-151.

No results published in the ECS summary.

Iowa

AI reading tutor, from summer 2025

A $3 million statewide purchase of a voice-based reading tutor for elementary schools.

Described as an investment, not a pilot.

New Mexico

District-led attendance pilot

Four districts used an AI tool that texts families when a student is marked absent.

No results published in the ECS summary.

Sources are the ECS summary by Sam Comai, June 24, 2025, and the Indiana Department of Education digital learning page, which hosts the final grant report. The Indiana survey drew 625 teacher responses, and the most common use was lesson planning, not tutoring. A survey tells you teachers liked it. It does not tell you a student learned to add fractions, which is what a principal needs to defend the renewal line.

Principal, math coach and teacher reviewing printed student work and a pilot calendar at a library table

02 The pilot plan

An eight-week AI tutoring pilot your board can follow

Long enough to get past the novelty, short enough that nobody forgets why it started. The setup week and the decision week are what most pilots skip, and they are where the yes or no comes from.

  1. Week 0

    Agree and clear

    Name the one skill gap you are testing, pick two to four teachers, finish the privacy review and write down the pass marks below.

  2. Week 1

    Baseline

    Each pilot class works five items on the target skill with the tutor. The cards give you a starting level for every student.

  3. Weeks 2 to 7

    Run on a schedule

    Two or three 15 to 20 minute tutor blocks a week, inside the period. Teachers regroup from the cards each Monday.

  4. Week 8

    Decide

    Rerun the baseline items, export the record to CSV, compare with a class that did not pilot and decide in one meeting.

03 Write these down in week 0

Five pass marks for an AI tutoring pilot

The numbers in the right column are our suggested starting points for a tutoring tool used inside class time. Change them to fit your schedule, but change them before the pilot, not after.

Use

Minutes per student per week, from the tool's own log

Median student reaches 40 minutes a week by week 3

Skill movement

Share of students who move from Not yet to Developing or Secure on the target skill

At least half of the students who started at Not yet

Teacher time

Minutes a teacher spends setting up and reading results each week

Under 30 minutes a week after week 2

Action

Times a teacher regrouped or retaught from what the tool reported

At least one regrouping a week per pilot class

Privacy

Data review complete, deletion terms in writing

Done before the first student session

The comparison class matters more than any single number. If the pilot classes move and a similar class taught the same unit without the tutor moves just as much, the tool did not earn its line. Coeducate gives you the skill levels for the pilot classes. The comparison class can take the same five items on paper in week 1 and week 8. If your district already runs MTSS, the same dated cards feed the eight-week review on the MTSS progress monitoring software page.

04 The pilot data

One student's week 1 card in a grade 5 fractions pilot

The item

Find 2/3 + 1/4 and explain your steps.

What the student typed

"3/7. I added the tops and the bottoms."

Multiply this by every student in four classes and you have the baseline. Rerun the same five items in week 8 and the cards say, student by student, whether the misconception is still there. That is the evidence the decision meeting needs, and nobody had to grade it.

Piloting for intervention groups rather than whole classes? Start from the math intervention software for middle school page or the reading comprehension intervention programs page.

Session evidence card, grade 5, adding fractions, pilot week 1

Fixed example

Diagnosis

Adds numerators and denominators straight across

The student added 2 and 1 for the numerator and 3 and 4 for the denominator. Asked whether 3/7 is bigger or smaller than 2/3, the student said smaller and was unsure why the sum of two positive numbers would shrink. The student has not linked the denominator to the size of the parts.

Next tutoring move / hint only

If you cut one pizza into thirds and another into fourths, are the slices the same size? What would you need to do before adding them?

Skill mastery read

Adding fractions with unlike denominators

Not yet
18%

Wrote 3/7 for 2/3 + 1/4 and described adding tops and bottoms.

Equivalent fractions

Developing
46%

Named 8/12 for 2/3 when prompted, not on their own.

Comparing fractions

Developing
50%

Saw that 3/7 is smaller than 2/3 but could not explain why that matters.

Explaining a strategy

Not yet
25%

Gave the procedure without a reason.

Teacher note

Equivalent fractions are close. Reteach common denominators with area models before more mixed practice, then rerun the same item in week 8.

Reteach in 5 minutes

Same-size parts first

  1. Draw 2/3 and 1/4 as area models on the same rectangle.
  2. Ask the student to cut both into twelfths and count the parts.
  3. Have the student write in their own words why the denominators must match before adding.

05 Before the first student logs in

Six questions to settle with any AI pilot vendor

Our own answer follows each one, including where we come up short.

  • What does the pilot cost, and what does year one cost after it? A pilot at no charge often hides the real price until the contract. Ours is the published price, so the pilot invoice is the forecast.
  • Is student work used to train models? Get the answer in writing. Ours is no.
  • Will you sign our state data privacy agreement? We have not yet signed a state DPA or joined a student-privacy consortium. Enterprise includes DPA review, which is where that starts.
  • What do we get out at the end? If the only export is a usage dashboard, you cannot judge learning. Ours exports every dated card to CSV.
  • What happens if we stop? Monthly plans cancel in two clicks in the billing portal, and you can ask for the workspace data to be deleted.
  • What will it not do? Ours does not plan lessons, grade tests, read handwriting or teach phonics and decoding. Do not pilot it for those.
Teacher with a clipboard checking a room of laptops set up for a pilot class

06 Pilot pricing

What an eight-week pilot costs at the published price

Two months of a monthly plan covers the full eight weeks. If the pilot passes, the annual plan is half the monthly rate, so the pilot does not lock you into the higher price.

Two teachers, Classroom plan

2 x $89 x 2 months

$356

Four teachers, Classroom plan

4 x $89 x 2 months

$712

Whole school, up to 600 students

$349 x 2 months

$698

If it passes, School plan for a year

annual billing

$2,088

Classroom covers one teacher and up to 60 students. District pilots across several schools can use the District plan with PO billing. Full terms are on the AI tutor for schools pricing page.

07 FAQ

Questions districts ask before an AI pilot

Writing the policy side first? The AI in schools page covers privacy and classroom rules, and the AI co-teacher page shows how a teacher runs the tutor alongside a small group.

What is an AI pilot program in schools?

An AI pilot program is a short, limited trial of one AI tool in a few classrooms before a school or district commits to a full purchase. A good pilot names the problem it is testing, fixes a start and end date, agrees in advance what result would justify buying, and collects that result from the pilot classes rather than from a survey alone.

How long should an AI pilot last in a school?

Six to eight weeks of actual use is long enough for most classroom AI tools, plus a setup week before and a decision week after. Shorter than four weeks and you mostly measure novelty. Longer than a semester and the pilot turns into an unpriced rollout that nobody formally approved. State pilots have run a full school year, but they tested many tools at once.

How much does an AI pilot cost for a school?

Many vendors offer pilots at no charge in exchange for a later contract, which makes the real price hard to see. Coeducate prices a pilot the same way as normal use. Two teachers on the Classroom plan for two months cost $356. A whole school on the School plan for two months costs $698, and the annual School plan is $2,088.

What should you measure in an AI pilot?

Measure the problem the tool is meant to solve, not general satisfaction. For a tutoring tool that means minutes of student use per week, how many students moved on the skill you targeted, how much teacher time it took, and whether teachers acted on what it told them. Agree on the decision rule before week one so the result cannot be argued after.

How do you run an AI pilot in a school?

Pick one problem and two to four willing teachers, get the privacy review done before any student logs in, take a baseline on the target skill, run the tool for six to eight weeks on a fixed schedule, then compare the pilot classes with the baseline and with a similar class that did not use it. Decide in a meeting with the numbers on the table.

Can federal Title funds pay for an AI tutoring pilot?

Often, yes, but it depends on the program and your district plan. Title IV, Part A allows spending on the effective use of technology, and Title I can fund supplemental instruction for eligible students. Your federal programs office decides what fits your plan. Indiana paid for its state AI pilot with federal pandemic relief funds.

Do students need accounts to pilot Coeducate?

Teachers create the workspace and add their classes. Students join the class the teacher sets up and work the items the teacher picks. The public demo on this page needs no account at all and stores nothing you type, so you can test the tutoring with a real problem before any student data is involved.

What happens to student data when the pilot ends?

Ask every vendor this in writing before the pilot starts. With Coeducate, student work is never used to train models, and a school that does not continue can ask for its workspace data to be deleted. The security page sets out what is stored, for how long, and where we stand on state data privacy agreements.

08 Pricing

Priced per program, not per seat

Annual billing saves 50%. There is no free tier, and the privacy posture is identical on every plan.

Classroom

$89/mo

One teacher, up to 60 students, CSV mastery export.

School

$349/mo

Up to 10 educators and 600 students, roles, SSO, scheduled reports.

District

$1,290/mo

Up to 75 educators and 5,000 students, standards mapping, SIS export, PO billing.

Full pricing and Enterprise

More AI tools for teachers

Product pages

Test the tutor on one item from your pilot unit

Paste a problem and a student's answer with their working, then read the card your pilot would collect. Nothing you type is stored.

Student work is never used to train models. No card required to try the demo.

Try the demo

Mastery evidence after every session.