How it works
AI tutoring platform that turns a session into evidence
A session, a diagnosis, an evidence card, a reteach plan. The teacher does none of the four steps and gets the output of all of them.
01 Step one
The student works, and gets coached
A student opens a problem from the material you already teach. They try it. When they stall, Coeducate does not produce the answer: it asks the next question that moves the reasoning forward.
That single default is what keeps the tool allowed in classrooms. A tutor that completes the work destroys the teacher's ability to trust any submission, which is why so many schools ended up banning general chat tools outright. See our classroom policy guide for the wording schools actually enforce.
Session transcript fragment
Student
I did 40 plus 20 percent which is 48, then 48 minus 20 is 28. That does not look right so I think it is just 40.
Coeducate, Socratic mode
You subtracted 20. Look at the second step again: 20 what? Twenty percent of $40 was $8. What is twenty percent of $48?
Student
Oh. It is 9.60 not 8. So it is 38.40, not 40.
Nothing in that exchange contained the final answer until the student produced it. That is the difference between a hint and a solution, and it is the whole reason the evidence further down is worth anything.
02 Step two
The engine names the misconception
"The student struggled with percentages" is not usable. It tells a teacher to reteach the entire topic to a class that mostly does not need it.
A misconception is narrower and far more useful: it says which specific move in the reasoning is wrong, and it quotes the student's own words as the evidence. That is what makes the record defensible when somebody asks how you know.
Not this
Struggled with percentages. Needs more practice with multi-step word problems.
Nothing to act on. Applies to half the class. Cannot be defended in a data meeting.
This
Applies the second percentage to the original base instead of the new one. Says "the 20s cancel out".
One sentence, one target, one five-minute fix, with the quote that proves it.
03 Step three
The evidence card is written
Three or four skills, each with a state, a mastery read and one line of evidence taken from what the student actually wrote. Underneath, a teacher note you can paste into a progress report without rewriting it.
Cards accumulate per student and per skill, so the second session's card is read against the first. That series is what student progress tracking is supposed to be, and it fills itself.
Anatomy of an evidence card
- Diagnosis
- The named misconception, plus two sentences on where in the work it shows.
- Next tutoring move
- The exact prompt the tutor uses next. In Socratic mode it is a question, never the answer.
- Skill mastery read
- Three or four skills, each with Secure, Developing or Not yet, a mastery percentage and an evidence line.
- Teacher note
- One sentence, report-ready, describing what the student can and cannot yet do.
- Reteach in 5 minutes
- One concrete activity with three steps, targeted at the diagnosed misconception.
04 Step four
Tomorrow morning has a plan
The card ends with an activity, not a recommendation to "review the topic". Across a class, cards group by misconception, so a teacher can see that eleven students share one fault line and four share another, and plan two small groups instead of one whole-class reteach.
That grouping is also what feeds the lesson plan generator: it starts from your class's actual misconceptions rather than a blank template.
Class view, one skill, 31 students
Grouped reteach. The six students in "Not yet" all share the same misconception about the base of a percentage. That is one small group and five minutes, not a lost lesson.
05 For the person who signs off
Two things a curriculum director asks first
Evidence you can present
Per-skill mastery per student, exportable, with the student's own reasoning attached. It survives a board slide and an MTSS meeting because the wording came from the student, not from a summary.
A privacy posture built for minors
Student work is not used to train models. No advertising, no data resale, deletable on request. Designed around FERPA and COPPA obligations, and our security page states plainly which formal attestations we do and do not hold.
Rollout, procurement and the full sign-off checklist are on the pricing page, alongside the cost-per-student estimator.
06 FAQ
What people ask after reading this page
Still stuck on something? Email app@coeducate.ai and a person answers, usually within one business day.
What does the student actually see during a session?
A work area and a tutor that responds to what they wrote. In Socratic mode the response is a question aimed at the specific step that broke, so the student is still the one doing the thinking. There is no separate "assessment mode" the student has to opt into, because the evidence is taken from the tutoring itself.
How long does one session take before there is a card?
A card is produced from a single piece of work, so a two-minute exchange is enough. There is no minimum number of turns and no session length to hit. A longer session simply produces more evidence lines against the same skills.
Where do the skill names come from?
From the subject, not from a generic taxonomy. Math sessions produce skills like "choosing the base of a percentage"; reading sessions produce skills like "inferring author stance". On the District tier those skills are additionally mapped to your own standards codes so the record drops into your reporting.
What if the diagnosis is wrong?
A teacher can override the state on any skill row, and the override is what exports. The evidence line quoting the student stays attached either way, which is the point: you are never asked to trust a label without the wording it came from.
Does the teacher have to read every card?
No. The cards aggregate into a class view grouped by skill, so the normal use is to open the two or three skills that are red across a section rather than 28 individual cards. Individual cards matter when you are writing a progress report or preparing an intervention meeting.
How does this fit a 45-minute lesson?
The common pattern is a 10-minute independent block where students work through practice with the tutor, and the teacher reads the aggregated skill view while it happens. The reteach suggestion is written to be a five-minute activity for the following day, not a new lesson plan.
More AI tools for teachers
See what a session actually produces
Run the Mastery Check on the sample work, or paste your own practice text. Nothing you type is stored.
Student work is never used to train models. No card required to try the demo.