Technology Should Draft, You Should Decide
If you read the first post here, you know I came to Terza from two directions at once, as a working cellist and as a software developer. That split is part of why this next principle isn't abstract to me.
Imagine the following scenario: you or your student is using an app to summarize lessons and it produces something that is wrong.
Not wildly wrong. The kind of wrong that's worse, actually: subtly wrong. The app produced something that sounds plausible but was not something that the teacher asked from the student in the room.
If that content had gone straight to a parent or a student, it would have been the teacher's name attached to it. Not the app's. Over time, mistakes like this can erode what you're trying to accomplish with the student and weaken your relationship with them.
When I realized this the core principle of this whole product stopped being a slogan and became, for me, the actual point of building it.
The industry default
Most AI tools built around recordings — lessons, meetings, calls — are built to remove friction. Record, generate, send. The fewer steps between the AI's output and the recipient's inbox, the better the product is assumed to be.
I understand the instinct. A review step is, by definition, more work than no review step. But I don't think that instinct holds up in a room where a teacher is being recorded teaching someone else's child. In that context, any content that is generated and shared appears to be from the teacher. And nobody should be put in the position of having said something they didn't say, to a parent who has no reason to doubt it came from them.
This isn't about some apps being more careless than others. It's a design choice, made early, by people who weren't necessarily picturing a professional's reputation riding on the output. Terza starts from the opposite assumption.
Terza generates. The teacher decides.
That's the governing principle, and I mean it literally, not as marketing language. Here's what it means mechanically:
Terza is a tool for teachers. The problem it is solving is to make it easier for teachers to get a full record of their students over time, leading to insights that would be otherwise impossible to see. The app can produce content for the student or parent, but that is initiated and controlled by the teacher.
After a lesson, Terza drafts observations and, when the teacher chooses, a practice note for parent or student. Neither of those goes anywhere on its own. The teacher reads it, edits anything that needs editing — a misattributed comment, an odd phrasing, a detail that needed more context — and only then decides whether to send it, and to whom.
Nothing reaches a parent that a teacher hasn't seen first. Not most things. Not the ones that seem obviously fine. All of them.
Why this is harder to build, not easier
For me, a review step is more product work than allowing an app to share the content it generates. It would have been simpler, from a build standpoint, to skip it.
The hard part isn't the gate itself — it's making review not feel like homework. A teacher finishing a full day of lessons doesn't want a new task waiting for them that feels heavier than just writing the note themselves would have been. If review is slow, or clunky, or requires re-reading a full transcript to catch one error, teachers will either stop doing it carefully or stop doing it at all — and either outcome defeats the purpose.
So the design problem I actually care about isn't "should there be a review step." That part isn't up for debate for me. It's "how do we make reviewing a draft take less energy than writing one from scratch, while still giving a teacher every reason to trust what they're approving." That's the harder, more interesting problem, and it's the one we're still actively shaping.
What it protects
It's tempting to describe this as a safety feature — a guardrail, a compliance nicety. I don't think that's the right frame. I think of it as professional protection.
A teacher's standing with a family is built over years, with the steady accumulation of small moments that create trust. A tool that puts words in their mouth — even well-intentioned, even mostly accurate — puts that standing at risk in a way that's completely outside the teacher's control. The review step is what keeps control where it belongs.
That's also what makes a tool like this something a teacher can actually trust.
Where I'd like your help
The principle isn't going to change. The experience of reviewing — how fast it is, what it looks like, what it should surface first — is still being built, and I'd genuinely like to hear from teachers on this. What would make reviewing a draft feel worth doing, rather than one more thing at the end of a long day?
I mean that as a genuine question, not a rhetorical one. If you teach and have thoughts on this, email me at hello@terza.app, I read everything.
I'll be writing more about how that review experience is taking shape as it comes together. For now: if an app is going to draft something in your voice, you should be the one who decides how and when it is shared.