iClicker AI Question Creator
Enables instructors to swiftly generate up to 50 unique questions tailored to their teaching style and course needs for polls, quizzes, assignments and more.
Project Overview
Problem
Instructors struggled to create fresh questions quickly, especially new teachers without existing materials and experienced instructors whose content had grown stale over time.
Process
Applied right-sized UX approach through progressive beta phases, focusing on learning over polish while maintaining educational product standards and pedagogical rigor.
Impact
Became the first AI feature to launch across the Macmillan Learning ecosystem with 71% of surveyed instructors reporting a positive experience.
“The AI Question Creator paves the way for a more dynamic, adaptive, and interactive learning environment. The tool’s intelligent design saves instructors time, allowing them to spend it where it’s most important, which is engaging with their students.”
— Steven Harper-Ray, VP of New Markets, Macmillan Learning
I helped shape the future vision of iClicker AI Question Creator, a product that scaled from a small beta of 5-10 instructors to a full release.
The goal was to provide a tool that would allow instructors to generate questions that worked with their teaching style.
I collaborated closely with the product owner, UX design director, and lead developer to define the vision of the experience and understand the application of AI in the core experience. With my role as the primary designer, I was responsible for leading conceptualization, refinement, and delivering design specifications for a final hand-off to the development team.
My Role:
Primary Designer
The Team:
UX Design Director
UX Designer (me)
UX Researcher
Product Owner
Lead Developer
Timeline:
Aug 2023 – Aug 2024
Problem
Instructors struggled to create fresh questions quickly, especially new teachers without existing materials and experienced instructors whose content had grown stale over time.
A lead developer from iClicker showcased a tech demo so I could understand the early functionality of iClicker AI Question Creator, and where there was room for improvements to the experience. The instructor could input a prompt for what topic the question would cover, and decide how many of each question type should be generated. The project was broken into three phases of work. Two betas where improvements are made iteratively based on instructor feedback, and a final release.
Phases of Work for the Project
Phase 1
Minimal Viable (5-10 Instructors)
Started with 2 days of UX work focused on removing usability barriers on the original demo rather than comprehensive design. The goal was to ensure the testing experience focused on validating the core AI functionality.
Phase 2
Expanded Beta (5,000 Instructors)
Scaled to a broader instructor base with iterative improvements to the experience based on the findings of the initial phase. Each phase had different technical requirements, especially around managing AI token costs.
Phase 3
Full Commercial Release
Launched across the entire iClicker instructor user base with confidence built through progressive validation.
Process
Mapping out the user flow for an instructor using the question creator allowed me to identify my open questions.
The original user flow I landed on with the product owner was complex and included direct deployment to courses as an activity that could be assigned as coursework. I proposed to scale the effort back so we could focus on the question creation experience so we can ensure the core experience using AI was sound for testing with users. Once I had approval from the product owner, I moved to drafting the first round of design improvements that would be used in phase 1.
Building out the full user journey, I was able to take a closer look at the question creation process and post question creation editing.
The goal was to make sure the AI Question Creator designs would integrate well with the present iClicker instructor web experience. I made use of existing design patterns to ensure consistency with the platform, and created new patterns where I needed to address a unique feature. Post question creation, if an instructor was unhappy with the questions created for them, they could choose to edit, flag, or delete their questions. The instructor needed to be in control of the content of their questions as they are the experts, and also be able to flag mistakes or inaccuracies so the team could make improvements. I was responsible for finalizing the first round of design specifications for the first beta.
Initial feedback showed instructor interest in the question creator, but also showed a need for an effortless question creation form.
To ensure instructors were confident in using the question creator, I iterated on the form design to provide clearer onboarding. Additional copy was integrated to provide instructors with a clear understanding of what the AI needed from them to generate their questions by providing the right amount of context. I also created sections in the form to help the instructor focus on individual parts as they progressively moved down the page to fine-tune their questions. This new form design was then utilized in the second phase with a larger beta of 5,000 instructors. As we received feedback live from instructors during the second beta, I made improvements to the onboarding through using Pendo to create Pendo Guides where I provided more information to instructors who might be new to prompting with an AI.
71% of instructors reported having a positive experience using iClicker AI Question Creator during its beta.
Instructors stated how they found the questions to be accurate and relevant to the question topics they used. In the times that they had to make edits, there were no issues reported in the question editing experience. One point of friction was the 50 question creation limit the development team had in place for the beta. This limit was in place so that the iClicker team could measure the cost of tokens used by the AI to create questions. With a clearer understanding of the cost of token usage, the team opted to remove the limit for the full release.
“Helpful. Would like to use it more!”
“It’s been great so far. This will save me a lot of time in class prep.”
“Excellent. Is this a permanent feature. Why is there only 50 question limit?”
Instructor quotes from Pendo survey
After the beta concluded, I continued to improve iClicker AI Question Creator using the instructor feedback.
The product owner wanted to include an option for instructors to use their own course materials to influence the AI on how to format their questions. This option was explained to be separate from just writing in a question topic as seen in the experience for the beta. I had looked at using an additional step to separate the two options so that the instructor had to make a deliberate choice. In the end, I choose to reduce scope in what developers would need to create and chose to integrate the choice of how instructors want to create their questions by placing it into the form.
Working closely with the development team, I documented flows for both the happy and sad paths the instructor might find themselves in.
The documentation I created was quite thorough for the scope of this project since we had a lot of paths to cover and didn’t want anything to go uncaught. We needed to be able catch invalid inputs, files that were the wrong type, files that exceeded the file size limit, files that were uploaded that were black, etc. As new information came in from the development team and the product owner, I was quick to make adjustments to our designs and documentation. After three phases of work, the iClicker AI Question Creator was integrated completely into the iClicker instructor web experience.
Impact
The iClicker AI Question Creator became the first AI feature to successfully launch across the entire Macmillan Learning ecosystem, launching as a beta to 5,000 instructors before full commercial release. Approximately 71% of instructors who shared feedback with the team reported having a positive experience utilizing the tool.
The beta-first approach proved faster than traditional upfront design methods would have been for AI features requiring real usage data. Strategic UX involvement from the beginning enabled smart go-to-market decisions without typical delays, while the lean collaboration model demonstrated that UX accelerates rather than delays AI experimentation when applied strategically.
Reflection
This was the first project I had the opportunity to take the lead on from initial conception to final release where I worked closely with the lead developer and the product owner. Some key lessons I learned from this project I’ve been able to apply to other projects I’ve worked on, such as:
- Prioritize testing the core experience of the product before designing extended features and deeper integrations. While iClicker AI Question Creator can in the future deploy activities, we needed to validate the core usage of AI first and see how instructors reacted to the experience before expanding on the concept.
- The importance of constant communication with the product owner and the development team as we worked on an experimental feature using AI. This was key to knowing when we needed to pivot on a design as more was learned about its capabilities.
- Organizing documentation into flows led to a better understanding on the intended behavior in user journey for both developers and quality assurance (QA).