Case study
What we learned building AI tools with Alberta schools
Honeybee Logic started with one school, one stubborn lesson-planning problem, and a decision to build with teachers rather than around them.
Who this is for
School leaders, educators, and builders working out where AI can be genuinely useful in day-to-day education.
It started with one school
Before Honeybee Logic was a company, it was a research collaboration with Connect Charter School in Calgary. We worked with teachers on a practical bottleneck: lesson planning takes a great deal of time, while generic chatbots lack the Alberta curriculum, the school’s approach, and the context behind a particular teacher’s lesson.
That project gave us something more useful than a speculative product idea. It gave us a real workflow, teachers willing to challenge our assumptions, and software we could improve in response to actual use. When we founded Honeybee Logic in July 2024, that work became the base of the Lesson Planning Assistant.
Why we started with lesson planning
Lesson planning is frequent, time-consuming, and close to the heart of a teacher’s work. It is also teacher-facing: we can make the planning process faster without asking an AI system to make decisions about a student.
The workflow is deliberately simple. A teacher starts with a curriculum objective, grade, subject, and instructional intent. The system produces a draft plan and supporting materials in the school’s context. The teacher edits, rejects, or reshapes the result. The output might become a worksheet, slide deck, activity, assessment, or rubric, but the teacher remains the author of the lesson.
Custom work became the product strategy
Our early customers came to us with different needs. Some wanted lesson planning. Others brought us reporting and school-specific workflow problems. We chose paid custom development while the product was still taking shape.
That choice shaped the business. A custom implementation solves one authority’s immediate problem, while the reusable pieces go back into the platform. The original Connect collaboration became the Lesson Planning Assistant. Other work contributed to report-card assistance and new planning features. During our first 21 months, custom development funded much of this learning, with subscriptions beginning to grow alongside it. The service work showed us what the product needed to become.
What the early traction tells us
By April 2026, Honeybee Logic had served 31 schools across four Alberta school authorities: charter, francophone, and Catholic systems in Calgary, Medicine Hat, Edmonton, and northern and west-central Alberta.
I am proud of that reach because it represents schools that paid for or deployed the work. It gives us evidence of demand, while the harder questions remain open. We still need better evidence on activation, repeat use, editing burden, time saved, willingness to pay, and the differences between selling to an authority and selling directly to a teacher.
What I have learned so far
Honeybee Logic has made my research and teaching more concrete. It is easy to talk about experimentation in the abstract. It is harder—and much more instructive—to decide what to build next when a teacher, a school leader, and a software team each see the problem differently.
- A narrow, recurring task is a better place to start than a broad promise to “transform education.”
- Local curriculum and school practices need to shape the product from the beginning.
- Teacher editing keeps professional judgment in the workflow.
- Paid custom work can reveal reusable product opportunities when we resist turning every request into a permanent one-off.
- Deployment and paid work show demand. Retention, learning quality, and measured time savings require their own evidence.
The next test
The next stage is less about adding impressive AI features and more about measuring ordinary use. Can a teacher get through onboarding without help? Do they finish a first plan? Do they come back? How much rewriting does a draft require? Which curriculum and pedagogy settings actually matter?
Those answers will determine whether Honeybee Logic should deepen its work with school authorities, expand a direct-to-teacher subscription, or change course. That is the honest state of the venture: real customers, a working product, encouraging traction, and important questions still open.
From evidence to action
Have a workflow worth examining?
Share the context, the people affected, and what you need to learn. The useful next step may be advice, a small experiment, or a focused build.
Discuss an applied-AI project