Drafting the report your teachers want to send

The fastest way to lose a teacher's trust is to put words in their mouth that they would never say. We learned that early. One of our first test reports described a seven-year-old's lesson as "a productive session with strong fundamentals." The teacher who read it laughed, then frowned. "Productive session." She doesn't talk like that. No teacher does. And if the report doesn't sound like her, every family on her roster will know in about two seconds that a machine wrote it, which defeats the entire point.

That report was technically accurate. It was also useless. The gap between an accurate summary and a report a teacher will actually send is wider than we expected, and closing it is most of the work behind lesson reports in IGNIFY.

The problem with "AI summaries"

If you have used a generic AI tool to summarize a meeting, you already know the three ways this goes wrong.

The first failure is generic. The summary could describe any lesson, any student, any day. "Worked on technique and made good progress." True of almost everything, useful to no one. A parent reads it once, learns nothing, and stops opening the next one.

The second failure is soulless. The model produces something grammatically perfect and emotionally dead, the prose equivalent of a stock photo. Parents can smell it, and teachers hate signing their name to it.

The third failure is worse, because it is confident and wrong. The model fills gaps with plausible invention: a piece the student did not touch, a skill they are actually struggling with. A wrong report quietly tells the family that nobody was paying attention, which is the opposite of what you are trying to communicate.

So the design question was never "can AI write a summary." It obviously can. The question was: what would it take to produce a report that a busy teacher, who knows the student and has strong opinions, would send with almost no changes? That is a much higher bar, and the only one that matters. A report the teacher rewrites from scratch is slower than no AI at all.

We set the target plainly: a teacher should be able to change one word and tap send. Not because every report will be perfect, but because that is the standard that forces the rest of the design to be honest.

Design principles, and the four-part model

A few principles fell out of chasing that standard.

IGNI is the editor. The teacher is the author. This organizes everything else. The model does not decide what the lesson meant. It drafts a structured first pass from what happened, and the teacher, who was in the room, owns the final word. We do not want a teacher rubber-stamping a machine's opinion. We want to save the teacher the blank-page tax of starting from nothing.

Structure beats fluency. A model that writes beautiful paragraphs will still wander if you do not tell it what a report is for. So we do not ask it for "a summary." We ask it to fill a specific four-part shape, and that shape is the actual product.

• Recap. What happened, in plain language a parent can follow. No jargon. This is the part that says "I see your kid, here is their hour."
• Report. What the teacher noticed technically. Bow hold, counting, breath control, whatever the craft of the thing is. This is where a teacher's expertise shows, and where a generic tool always goes vague.
• Memory. What to practice and hold onto before the next lesson. Concrete and small. This is the part families actually use during the week.
• Story. Where it is heading. The recital in spring, the next level, the reason this week's small thing connects to something bigger. This is what turns a status update into a reason to stay.

The order is deliberate. Recap earns attention, Report shows expertise, Memory makes the week useful, Story gives the family a future to keep paying for. Drop any one and the report tilts back toward being a notification.

Plain language is not dumbing down. The Recap is written for a parent who does not play the instrument and does not want a lecture. The Report can be precise and technical, because that is what proves a real person was teaching. Keeping those two registers separate, instead of blending them into mush, was one of the more important things we got right.

The teacher stays in control, structurally, not as a courtesy. Nothing reaches a family until a teacher approves it. The draft is editable everywhere. Recordings, when a studio uses them, are consent-based at the family level, and customer content never trains outside models. None of that is a marketing line. It is the only arrangement under which a teacher should be willing to attach their name.

What surprised us

We got several things wrong early, and the corrections taught us more than the wins.

We thought longer reports would feel more valuable. They did not. Our first drafts tried to be thorough and ended up exhausting. Parents skim. The reports that landed were short, specific, and finished in under a minute of reading. We cut hard, and the four-part shape is partly a forcing function to keep each section tight.

We thought praise was the safe default. It is not. Relentless positivity reads as fake, and it erodes the credibility of the praise that is real. A report that can name a genuine struggle, gently and usefully, is the one a parent believes when it later says the kid did something well. Honesty is what makes the encouragement worth anything.

We underestimated how much the teacher's voice matters and how little it costs to keep. Early on we optimized for a clean, neutral house style, and teachers found it flattening. The fix was not complicated: draft in a plainer register, leave room for the teacher's own phrasing, and never reach for words a teacher would not use out loud. "Productive session" is gone.

The biggest surprise was strategic, not technical. We started treating lesson reports as a communication feature. We now think they are the retention engine. A family that gets a short, real, weekly note about their child's progress understands what they are paying for. They feel seen. They renew, and they tell other parents. The report is where the value of the lessons becomes visible, and visible value is what keeps a studio full.

The short version

A report your teacher will actually send is not a summary with better grammar. It is a small, structured, honest piece of writing that sounds like the person who taught the lesson, and IGNI's job is to get it most of the way there so the teacher can finish it in a minute. Get that right and you are not adding another notification to a family's pile. You are giving them the one message they open every week. That is the report we are trying to draft, every lesson, for every teacher.