The AI Classroom Companion
designed for the learner’s need, not the technology: help that’s easy to find, honest, and never costs you your place
Bottom line
I lead the experience strategy for our first AI learning assistant in the classroom. Before any vendor tool could shape the experience, I set the standard it has to meet.
It stands on a year of ecosystem research: learner interviews, a 23,000-case friction review and journey mapping across the whole student experience. This is where that research meets the classroom.
One plain label. Help without losing your place. Honesty about what it knows. A clear door to a human. Then I built the enablement kit and the governance so faculty, support teams and leaders can adopt it safely. It goes live in a market test this October.
~1%of 23,000 support cases came from the classroom. Not quiet: no door.
4 → 1four names for help became one plain label
8experience standards any AI tool gets scored against
RoleLead Experience Strategist
experience strategy, enablement, AI governance
TimelineNov 2025 → now · research → standard → market test
TeamWorkforce + academic leads, IT, support, faculty, vendor
SkillsAI experience design, service design, knowledge architecture, governance
The board
problem → left to right · fix ← right to left
Help hid in plain sight
It sat right on the page, but under four different names and no label saying “this is help.”
I’m trying to understand this unit at 10 pm. But I can’t tell which button is help. I feel like asking for help is a puzzle, too.
Too many steps to ask
Before a first question: connect an account, pick an agent, read a welcome about the product, choose a path.
- connect account STEP 1
- pick an agent STEP 2
- read the product welcome STEP 3
- choose a path STEP 4
friction dressed as guidance
No door to a human
The classroom had no escalation path, so its problems never reached anyone. A dead zone, not a problem-free zone.
“I need human help.”— a learner need, from the journey work
Here’s how we made help easy to trust.
design around the technology → design for the learner’s need, then hold any tool to it
Name it for the job
I recommended a plain label over a mascot. A name adds a step: first you learn that the name means help.
- before
product name · tutor · sage · persona - afterGet Learning Help
help with your coursework, anywhere in your classroom
Set the standard first
Eight tool-agnostic standards every support surface inherits, so we hold the tool to the experience, not the other way around.
- Find help without leaving the lesson
- It already knows where you are
- Go back to your exact spot
- One name, used everywhere
- Honest about what it knows
- A human in one step
- Accessible and text-first
- Measured from day one
Get every team ready
An enablement kit with one path per role, and one question-card pattern that people and the AI assistant can both read.
- TEACH · faculty
- SUPPORT · support teams
- BUILD · builders
- GOVERN · leaders
Can my students trust what the AI tutor tells them?
It answers from your course content and says so when it isn’t sure. Your guidance comes first: if it differs from you, students go with you and flag it.
How I orchestrate it
People
- listened to learners first: what they need in the moment
- aligned faculty, support, IT and leaders on one standard
Product software + the devices it lives on
- designed the label, placement and first question
- built the knowledge architecture and the path to a human
Process
- designed for the need first, then scored each tool against it
- governed with an intake sheet and an inventory of every AI agent
The ripple
- Learners help in the moment, without losing their place
- Faculty answers to their questions before launch
- Support handoffs with context, capacity protected
- The institution a template for every AI tool that comes next
Where it stands
- ✓Ecosystem research + friction review
- ✓Current-state journey map
- ✓Experience standard v0.2
- ✓Enablement kit + governance templates
- →Market test, Oct–Dec 2026
- →Test the label with learners
No outcomes claimed yet. I wrote down what it takes to keep it running, so we scale on purpose.
Illustrative reconstruction · simplified flows · no internal systems or vendor names shown