Zero-to-one AI projects: AI-assisted grading and onboarding
BrainPOP Assisted Grading for CER powered by AI. Highlights from the AI project Sarah Mondestin worked on with the BrainPOP Science team.
Part 1: The situation
At BrainPOP, I worked as a content designer with the Science product team and conducted extensive user research with teachers and students in the classroom. While teachers genuinely loved our new science investigations and their impact on student learning, a critical pain point emerged: grading was tedious and time-consuming. My direct observations in classrooms confirmed two major issues:
The overall product onboarding for the science platform was confusing.
The grading process was a burden.
We recognized an opportunity to leverage AI to alleviate this grading pain. However, AI-assisted tools were relatively new concepts for many teachers at the time. In user research, teachers expressed apprehension and a need for significant guidance and reassurance to trust AI tools.
Part 2: The task
I collaborated in an Agile environment with learning designers, product designers, user researchers, engineers, and product and project managers,
My primary tasks were to:
Create voice and tone guidelines as a “north star” for content.
Simplify the overall onboarding experience for new teachers and users.
Design the content support for a new AI-powered grading tool, using a large language model (LLM) to provide grading feedback comments and scores, to make grading easier and faster for teachers. This involved ensuring teachers felt comfortable and in control of using AI to assist with their grading.
Part 3: The action
What I did
My primary role on the AI-powered grading tool was to be the voice of the user and the advocate for content clarity. The product was a 0-1 initiative using an LLM to provide feedback to students.
My main challenge was to make sure users understood how it worked and felt they could trust its output. I developed a content strategy that included a voice and tone chart, in-product copy guides, help articles, and prompt engineering.
What I did
Developed a voice and tone chart to guide content by leading content workshops with team members and stakeholders. In order to create content-first designs for new AI features, I led voice and tone workshops to ensure we aligned on our product personality.
What I did
Created all the in-product copy that explained the tool's purpose and guided educators through the workflow, including microcopy for error states and AI-assisted grading toggles. This created a type of “guided interaction” throughout the product experience instead of a front-loaded onboarding with too heavy a cognitive load.
What I did
Prompt engineering: A key part of this project was collaborating with learning designers to translate our pedagogical standards into the grading rubrics that we used to feed the LLM, a form of prompt engineering, to ensure the feedback was consistent and accurate.
What I did
Developed a detailed help article that demystified the AI's functionality. Users trusted us but not the new tool. Working with engineers and learning designers, I created a simple help article to explain AI-assisted grading and LLMs to users in a way they could easily understand.
What I did
Simplified the onboarding experience for users. We were preparing users to do something brand new and interact with tools they weren’t familiar with. Smooth and simple onboarding was important.
Part 4: The result
The impact of these UX content design efforts addressed the problems identified in our research:
Increased teacher comfort and adoption of AI: Teachers felt significantly more comfortable using the AI tool to assist with grading, leading to its successful adoption. It genuinely made their grading process easier and more efficient, reducing a major pain point.
Reduced cognitive load and user frustration: The simplified onboarding and clear, concise language throughout the product reduced cognitive load and minimized user frustration, making the entire BrainPOP Science platform easier and more enjoyable to use.
Decreased support inquiries: As a direct result of clearer onboarding and comprehensive, plain-language explanations for the AI tool, we observed a noticeable reduction in support inquiries related to platform usage and AI feature understanding, saving the company valuable time and resources.
UX content design strategically bridged the gap between complex technology (AI/LLMs) and real-world user needs (teachers’ grading), leading to greater efficiency, ease of use, and overall user satisfaction.
