Reflection
What it means to design AI for teachers
A first-person look back at building the AI Teacher Assistant Platform — what I learned about designing AI products for educators, why the human stays in the loop, and what this work says about leading AI innovation in education.
Designing AI products for educators
When I started this project, I resisted the temptation to build "a chatbot for teachers." Teachers don't need another open text box — they need their time back. So I designed around the work itself, not around the technology. Every decision flowed from one question: does this give a teacher a stronger starting point, faster, without adding cognitive load? That reframing — from "AI tool" to "productivity platform for a specific professional" — was the most important thing I did.
Why human-in-the-loop matters
The fastest way to lose a teacher's trust is to make AI feel like it's making decisions about children. I built the opposite stance into the product: AI proposes, the teacher decides. Outputs are visibly labeled as drafts, nothing sends automatically, and every generation carries a reminder to review for accuracy, bias, and fit. This isn't a disclaimer — it's the design. Keeping the human in the loop is what makes the automation trustworthy enough to actually use.
Prompt engineering & workflow automation as instructional design
I came to see prompt engineering as a form of instructional design. Designing a reliable generation template — deciding which inputs to collect, how to structure an output, how to bake in standards alignment and Bloom's/DOK balance — is the same discipline as designing a strong lesson: clear objectives, scaffolded structure, and predictable quality. By moving that expertise into the workflow, I let teachers benefit from good prompting without ever having to learn it. The instructional design is invisible, but it's everywhere in the product.
UX for busy teachers
Teachers work in stolen minutes — between classes, during a prep period, late at night. That constraint drove the UX more than any aesthetic preference. A command palette so nothing is more than two keystrokes away. A consistent form-to-draft pattern so every tool feels familiar after the first. Dark mode for the after-hours reality of the job. Calm visuals and clear trust cues so the product never feels like one more thing to manage. Respecting a teacher's attention is the user experience.
Leading AI innovation & building viable edtech
This build is, for me, a demonstration of range: I can analyze a real professional workflow, translate instructional-design expertise into product logic, make principled choices about responsible AI, and package it all as something a school or district could actually adopt. It shows I can sit between educators and engineers — speaking both languages — and lead AI initiatives that are pedagogically sound, ethically grounded, and commercially credible.
What it demonstrates
This project demonstrates my ability to lead AI innovation in education end-to-end: identifying a high-value problem, designing human-centered AI workflows, embedding responsible-use governance, crafting a polished SaaS-grade experience, and articulating an implementation and ROI case to decision-makers. It is evidence that I can build commercially viable edtech without compromising the professional judgment of the educators it serves.
What I'd do next
If I carried this forward, I'd connect a real model behind the existing workflows, add voice and speech-to-text for hands-free use, and integrate with Google Classroom, Canvas, and Teams so drafts publish where teachers already work. I'd run a measured pilot to validate the time-savings claims against a real baseline, and I'd build adaptive recommendations that learn from each teacher's edits. The architecture was designed for exactly that next step — the demonstration is the foundation, not the ceiling.