We opened Teacher's Buddy to a small group of K-12 teachers before public launch, and we made a deliberate decision to stay in close contact with them for the duration of the early-access period. Not a survey at the end. Weekly check-ins, a shared feedback channel, and in several cases, screen-share sessions where a teacher walked us through how they were actually using the tool.
Forty teachers across four states. Elementary, middle school, and high school. Subjects ranging from 2nd-grade general education to 10th-grade AP Chemistry. The early-access period ran for six months. This is what we learned.
What Teachers Actually Used Every Week
We expected lesson planning to be the primary use case. It was, but not in the way we assumed. Most teachers did not use Teacher's Buddy to generate lesson plans they had never written before. They used it to produce a first draft of lesson plans they had written before, typically because they were teaching a unit again and wanted a structured starting point without copying last year's notes verbatim.
The phrase we heard most often in the first month was "it saves me the blank-page problem." That is a real thing. The hardest part of writing a lesson plan from memory is getting the structure down without staring at an empty template for 15 minutes. A draft that is 70% right is much easier to edit than a blank page.
The second highest-use feature was sub plan drafting. This surprised us a little, because sub plans were a relatively late addition to the initial build. What we had underestimated was how often teachers find themselves writing sub plans in a hurry: the evening before, an early morning emergency. A tool that could take a brief description of what the class was working on and produce a complete, readable sub plan in a few minutes turned out to be genuinely valuable in a way that routine lesson planning was not. Routine planning has a week of lead time. Sub plan writing often does not.
Parent communication drafts were the third most-used feature among teachers who tried them. The adoption pattern was interesting: teachers who started using parent email drafts in the first two weeks tended to keep using them. Teachers who did not try them early rarely tried them at all. The barrier seemed to be the first attempt, not the ongoing use.
What We Changed Because of Feedback
The first significant change came in month two. Several teachers told us that the initial lesson plan outputs were well-structured but used generic examples that did not match their specific content. A 4th-grade teacher in Texas and a 4th-grade teacher in Minnesota might be teaching the same math standard, but their classroom context, their students' prior knowledge, and even the vocabulary they use for certain concepts can differ. Generic examples produced plans that felt complete but required heavy editing to feel like theirs.
We added a context field that lets teachers describe their class before generating a draft: current unit, where students are in the content sequence, any particular gaps or strengths in the room. The improvement in output quality was immediate, and teachers who used it reported spending significantly less time editing. The lesson for us was that our initial assumption, that teachers would add context in the edit phase, was wrong. Context before generation produces a better starting point than context added after.
The second change came from a consistent complaint about the assessment section of lesson plan drafts. Several teachers in the group told us the formative assessment suggestions were generic, and that they typically deleted the assessment section and wrote their own. We looked at the outputs they were deleting and they were right. The suggestions were structurally correct but not tied to the specific lesson objective. We rebuilt the assessment generation to take the lesson objective as an explicit input and produce checkpoint questions that were aligned to it. This was a meaningful improvement that came directly from teachers telling us something was not useful.
The Feedback We Did Not Expect
A few teachers told us they felt uncomfortable with how good the drafts were. Not a complaint, exactly. More of an honest expression that they were not sure how much of the plan was theirs when they used a draft that required relatively little editing.
We talked about this a lot internally. The teachers who expressed this concern tended to be the ones who cared most about their craft, which is not a coincidence. They had built an identity around being thoughtful planners, and using a tool that handled the structure efficiently felt like it was changing what that identity meant.
We do not have a complete answer to this. What we said to those teachers, and what we believe, is that planning quality is not measured by how long the plan took to produce. It is measured by how well the plan serves students. A teacher who uses a draft and edits it thoughtfully, who knows why every piece is there and what they would change if the class responded differently, is planning well. A teacher who writes from scratch but produces a plan that is structurally identical to last week's is not planning well. The tool does not define the quality of the thinking.
What Surprised Us About Who Used It Most
We expected early-career teachers to be the highest-use group. They have the most to gain from structural scaffolding and typically have less experience building lesson frameworks from memory. That assumption held up in the first month. After month two, a different pattern emerged.
The highest sustained users were mid-career teachers: five to twelve years of experience, established in their classrooms, responsible for teaching multiple preps or multiple sections of the same course. These teachers had the experience to evaluate an AI draft quickly and the context to edit it specifically. They were also the teachers who most often described their prep time as an actual constraint, not a professional development opportunity.
Early-career teachers often wanted more explanation of why a lesson was structured a certain way, not just a completed plan. That was useful feedback that pointed toward a future feature rather than a problem with the current one.
Where the Tool Fell Short During Early Access
There is one area where early-access feedback was consistently critical, and where we are still working on improvement: differentiation. The standard lesson plan draft does not produce meaningfully differentiated versions of the same lesson. Teachers with identified students on IEPs or 504 plans found the differentiation suggestions generic to the point of not being useful. "Provide additional scaffolding" is not differentiation. It is a placeholder.
We have made some progress here, but the differentiation problem is genuinely hard. Good differentiation requires specific knowledge of which students, which learning needs, and which parts of the lesson are the sticking points. That information is complex to input and complex to reason about. We are still learning from special education teachers in our user group about what would actually help versus what just adds a well-formatted section to the document that no one reads.
This is not an argument to lower expectations for what a planning tool can do for diverse learners. It is an honest statement that we have more to build in this area and that we are doing it in conversation with the teachers who need it most.
What We Built Next Because of the Feedback
Two features came directly from early-access requests. The first was the sub plan follow-up email: after you write a sub plan, Teacher's Buddy can generate a brief message to parents explaining that you were absent and the class covered a specific topic. Several teachers told us that notifying families after a substitute day was something they always meant to do and rarely found time for. The email draft takes 30 seconds to produce and two minutes to edit.
The second feature was a unit-level planning view that lets teachers see a multi-week sequence before drilling into individual lesson plans. Several teachers told us they wanted to do the backward-design work at the unit level, then generate lesson plans within that structure, rather than planning lesson by lesson and hoping it added up to a coherent unit. We built the unit planner for the public launch because of that feedback.
To the 40 teachers who used the tool through its rough early form and told us honestly what was not working: that feedback is why the current version is better than the one you started with. We are still building, and the people whose judgment we trust most are the ones who are in a classroom every day.