01 / Personalization without the workload
Students need plans that fit them. Teachers should not have to build a personalized plan from scratch for every student and assignment.
Turning student performance data into teacher-led next steps
client workIAUser researchPrototypingSystems thinking
Summary
IntelliMind is an AI-assisted learning platform and study app developed by Minerva Next. It connects learning materials, schedules review, supports visual recall, tracks progress, and recommends what to study next.
Our client engagement focused on its educator dashboard: helping teachers manage classes and assignments and understand student progress. I led UX alongside a product manager and a UI designer, then independently revisited the workflow after delivery.
The problem
Teachers could see the data, but not the next step.
After delivery, usability feedback exposed a gap: users were unsure what to do next after opening the educator dashboard. A teacher’s question made the problem concrete, the interface could show performance without making a response clear.
Context
We delivered the dashboard, but a teacher still asked what to do next, so i refuse to confuse done with good.
We started with a feature list rather than a defined end-to-end workflow. The delivered dashboard emphasized charts and visual reports, leaving teachers to interpret the data, identify who needed attention, and decide how to respond. That added mental effort between seeing a signal and taking action.
Insights / Independent redesign
I mapped the missing workflow before reorganizing the screens.
I connected the work teachers set, the signals they monitor, and the responses they choose into five steps. The redesign focuses on the handoff from noticing a need to preparing and reviewing support.
- 01
Create & assign
Set the work and expectations.
- 02
Track
Review submissions and performance.
- 03
Identify
Check who needs support and why.
- 04
Prepare support
Nudge, or draft a targeted plan.
- 05
Review & deliver
Adjust and approve what students receive.
Research
I revisited educator needs before reorganizing the screens.
Five educator interviews informed the project. For the independent redesign, I revisited that research and mapped sample scenarios to connect personalization, actionable feedback, and teacher control to the workflow.
02 / Support that fits the student
A student’s performance is only part of the picture. Teachers need to decide what support fits their circumstances.
03 / A concrete next step
A clear next step and targeted remedial work give teachers a way to respond and students a way forward.
04 / Teachers in the loop
AI can prepare the options. Teachers decide who needs support, adjust the plan, and approve what goes out.
1 / 4
Design principles
Make the next step useful, contextual, and teacher-led.
- A concrete next step
Connect learning signals to support teachers can act on.
- Support that fits the student
Keep the teacher’s understanding of each student’s circumstances in the decision.
- Personalization without the workload
Let AI prepare options; teachers adjust and approve what goes out.
Design decisions
Turning learning insights into action while keeping teachers stay in the driver's seat
I organized the redesign around three teacher decisions: understand the situation, choose who needs support, and review what to send. AI supports preparation; the teacher retains approval.
Teacher authority
Decide what fits
Check the situation and recipients
Choose whether a reminder or learning support is appropriate for each student.
Adjust and approve the plan
Review the proposed support and decide what to send. AI generation is not approval.
AI assistance
Prepare the next step
Up to 3 students → suggest a nudge
For a small number of missing submissions, prompt a reminder to submit.
More than 3 → prompt a support review
Invite the teacher to check for a shared difficulty and create a study plan if appropriate.
Draft personalized remedial work
Use student work to prepare activities the teacher can adapt, rather than starting from scratch.
01 / Understand the situation
Real-time visibility into every classroom to be in the know
The 2-tab filters shows you who's engaged and who needs help, across every classroom, in real time.
Design choicethe filtering feature would allow teacher to monitor and review student performance by class and by assignment, which preserves the teacher’s knowledge of each student’s situation as opposed to turning a status into an automatic bulk intervention.
See the decision in action
Start with the question, then inspect the relevant work
The recording demonstrates the two-step filtering structure.
02 / Choose who needs support
Students are grouped by submission status so teachers can pull the right students together for targeted instruction.
The submission status flags who needs help right now, so teachers can act while students are still working and students get support before they fall behind.
Design choiceProviding color-coded submission status would help teachers identify gaps in understanding that may have gone unnoticed during a regular class. In this way, teachers see where students stand, and know exactly who needs more support before the next lesson.
1Nudge and plan actions sit above the list, while individual recipient selection is not visible.
See the decision in action
Review a group before sending one nudge
The recording shows the outreach workflow following student selection.
03 / Review the next step
Putting AI in charge of personalized plans while teachers making the final call for content boundary
AI would use coursework evidence and teacher-confirmed needs to adjust activity complexity for cohorts facing similar difficulties, then draft study plans that fit. Submission issues would prompt a teacher check, rather than determine a learner’s ability on their own.
Design choiceAI would prepare appropriately challenging activities for each cohort across subjects; teachers would inspect, review, and adjust the plans before delivery. The aim is to make personalization manageable without multiplying preparation work or contributing to burnout.
1A brief recommendation names resources, but does not show a timed activity sequence.
What still needs testing: whether reviewing and editing the proposed plan reduces preparation time without adding an equally demanding checking task.
See the decision in action
Inspect the plan before it goes out
The recording shows how teacher review fits into the support workflow.
Reflection
A few things that I learned
The independent redesign connects signals, student selection, and teacher-reviewed support in a high-fidelity prototype. It is not evidence of a shipped integration or measured time saving.
A polished screen still needs a clear next step.
A polished screen is not a good screen if users don't know what to do with it. When a feature list leaves the workflow undefined, mapping what people need to see, decide, and do is part of the design work, not a step to wait for someone else to finish.
Review should save work, not simply move it.
I would compare the effort of preparing support from scratch with reviewing an AI draft: review time, number of edits, and whether the final activities fit the student. Faster generation is not enough if checking takes equally long.
Let teachers calibrate a plan, not rewrite it.
Next explorations would let teachers set difficulty, a time budget, and a deadline before regenerating relevant activities. These controls remain proposals to test, not validated improvements.
That's the end of this story. If you're still curious, here are two more.