Replacing a 5-day specialist-led charger activation workflow with a 4-minute guided wizard.

ChargePoint sells EV charging hardware and runs one of the world’s largest charging networks: over 1.3 million charging ports. Every day a station isn’t activated is a day it’s not earning.
Activation used to take 5 days and a dedicated specialist. I designed a guided wizard any organisation admin can run in 4 minutes.
- 4 minutesBeta
- activation time, down from 5 days
- 4 iterations in 1 year
- each tested before the next began
- 40+ defects caught
- pre-release through embedded design QA on dev builds
- Role
- UX/UI Designer (Design Lead) End-to-end design, user research, design QA on dev builds, user testing, prototyping
- Team
- 5 Engineers, 1 PM, 1 Customer Experience Specialist
- Timeline
- Apr 25– Apr 26
- Interactive AI Concept
- View

Context
How activation worked before
Electricians filled out a 3-page paper form by hand and emailed it to CX.

Activation form
The 3-page paper form electricians filled out and emailed to CX.
- 1.
- 2.
- 3.
Three legacy platforms, one new suite
Activation logic was split across three legacy platforms. Polaris Suite is the new enterprise platform that replaces all three.
The people involved

- 1. CX agents (Customer experience)
- Receives the paper form and coordinates between org admins, field teams and the platform.
- (secondary persona)
- 2. Deployment specialists
- Configures stations across 3 legacy platforms. The only ones who can activate.
- (secondary persona)
- 3. Org admin
- Customer site manager. Wants instant self-serve setup without relying on specialists.
- (primary persona)
Problem
- Expensive by default
- Every activation required a specialist, costing an estimated 2,000+ specialist hours per year.
- Three systems, one task
- No shared flow across the three legacy platforms.
- No bulk activation
- 10 stations meant running the wizard 10 times.
- Expert-only settings
- Most admins couldn’t choose the right plan or policy alone.
How do you let a non-expert activate charging stations on their own, whether it’s one or a hundred?
What I walked into
I joined weeks before the MVP shipped. No time for research. A two-day dev workshop to understand the system, then ship.
01 The Wizard (MVP)
A three-step wizard for the simplest case.
What I built

The banner surfaces stations ready to activate. 1.Charger Management banner
Notification that stations are ready for activation in the suite.
2.Ready-for-Activation
List grouped by model family and site.
3.A wizard (Org & Plan → Energy Management → Summary)
A three-step wizard with plans, warranties and policies pre-selected. Most admins confirm the defaults and move on.
Real research started after MVP
Insight
30+ interviews with org admins, deployment specialists, and CX. The finding that shaped everything after: admins didn’t need a faster wizard. They needed to know when something was wrong and what to do about it.
Two products, one handoff
I also designed the Installer App. That meant I could close the gap between field and platform myself: data captured by the technician on site flows into Polaris Suite, so the admin starts with a station that is already pre-filled.

02 Bulk and recovery
What I built
1.Copy/Import Configuration
Large activations repeat the same setup, so admins can reuse an activated station’s settings across new ones.
2.Token (license) selection
Every station needs a token (license) to go online. The right one is now pre-selected. When an admin needs to override, they see which token fits before choosing.
3.Token-mismatch recovery
Incase wrong or expired token (license) selected, the system failed activation. The admin had to call support. Now the system tells them what went wrong and how to fix it in place.
03 Pre-activation checks
Admins could finish the entire wizard and only then discover the station couldn’t activate. Once the backend supported it, the system now checks before they start.
What I built
1.Blockers caught early
(1) pending GPS/ location co-ordinates of the station (pinpoint), (2) missing gateways, (3) unsupported hardware. Surfaced before the wizard.
2.Activation progress
Before: the admin pressed Activate and waited. No feedback, no way to tell if it worked, failed, or was still running.
Now: per-station status with success, failure, and in-progress, a resolve button on failures, and a progress bar that persists across page reloads.
The DC blocker banner alone saves us a support ticket every other day.
04 Clusters
A charging station is called a cluster when it has 10+ outlets (dispensers) sharing one charger ID. Instead of listing each separately, the hierarchy groups them under one parent so the admin sees one station.


Staying in the build
From iteration 3 on, I ran design QA on every dev build myself: clicking every state, breaking the flow on purpose, logging defects in Jira. 20+ defects caught pre-release per iteration, before they reached the users.
How AI fits in
In the process
Copilot for summarising 30+ interview notes and drafting specs.
As a concept
Copy Config only works if you know which station details to copy. The AI concept finds the org’s past setup for matching hardware and suggests plans, warranties and policies. When there’s no history, it shows all options instead of guessing.
What I learned
Stay close to support
The most valuable feedback came from CX telling me what annoyed them that week, not from formal UAT.

The unhappy path is the product
Designing failures as first-class moments did more for trust than any polish on the happy path.








