Case Study · Take-home Assignment · Intuit
This is a learning strategy brief and job aid for Mailchimp Analytics AI, built as a take-home assessment for an Instructional Designer role at Intuit Academy. Not internal Intuit work, and the outcome of that application isn't part of this write-up.
The Problem
The brief: Mailchimp is shipping Analytics AI, and Intuit's seasonal Tier 1 support agents need to field live customer questions about it starting on day one. These agents already know Mailchimp cold, so the gap isn't the product, it's this one feature. Two constraints shaped everything after that: they're on live chats and calls most of the day, and a seasonal ramp has no room for a course that takes them off the floor.
Modality
A layered path, not a single course: a 60-second video tour, then guided sandbox scenarios inside a downloadable job aid, then an FAQ for point-of-need answers. The video front-loads the value props in under a minute. The job aid is pull, not push, it's there when an agent actually needs it, mid-call.
Cognitive ladder
Each modality targets a different level of Bloom's taxonomy on purpose. The video builds understanding of the marketing team's key messages. The sandbox scenarios push into applying, analyzing, and evaluating, agents run real prompts and read real output. The FAQ supports recall and transfer in the moment a customer asks.
Speed to competency
Chunked and sequenced this way, a seasonal agent gets to "ready to help a customer" fast, then deepens through hands-on practice, without a long stretch away from the phones.
Artifacts
The cover, the reasoning behind the sequence, a sandbox scenario, and the FAQ agents keep open while they talk.
The job aid's cover states exactly who it's for and how to use it: skim the value props before a call, run the sandbox to practice, keep the FAQ open while chatting.
One path, three modalities, each one raising the level of thinking required.
Agents type the actual prompts into a hypothetical sandbox, read what Analytics AI would say back, then answer reflection questions built to catch the moments a customer would actually push back.
The reference an agent keeps open mid-call, including the honest one: responses are AI-generated and may not always be accurate, so nothing sends, creates, or changes without the customer confirming first.
Measurement Plan
Engagement
Video completion, sandbox scenarios run, and drop-off between the video and the job aid
Proficiency
A split test of trained versus untrained agents on customer NPS and Analytics AI adoption
Business impact
Product-support ticket volume, to confirm the training holds up under real call load