My team wrote certification exam questions by hand, and the company's position on AI was still forming. Any AI use had to hold up to accuracy standards for a certification exam.
Bringing AI into exam development
My team wrote certification exam questions by hand while the company's position on AI was still forming. I piloted AI-assisted drafting with a human check on every question, and quiz-writing time fell by roughly 90%.
- Role
- Led the pilot as Team Lead
- When
- 2025
- Scope
- My team's exam and content work
- Worked with
- Peer team leads
Less time spent writing exam questions
Less time investigating class inconsistencies
The situation
What I did
- Piloted AI-assisted exam question drafting with Microsoft Copilot. Every drafted item was checked against the source content before use.
- Built department-level responsible-use practices with peer team leads, covering hallucination risk and prompt writing.
- Applied Copilot to investigating inconsistencies between classes.
Result
Quiz-writing time dropped by roughly 90%, and the workflow became standard process for the team. Time spent investigating class inconsistencies fell by 80% to 90%.
What I'd do differently
I'd track how AI-drafted questions performed against hand-written ones once they were in live exams, such as how often learners missed them and whether any got flagged. We checked every item for accuracy before use, but I didn't measure whether the drafted questions tested the material as well as the ones we wrote ourselves.
Building or scaling customer education? Let's talk.
I'm looking for leadership or builder roles in customer education, onboarding, and enablement, where I'd be creating something new or close to it. I'm especially drawn to SaaS companies where education sits close to product.