Accenture Wildlight
The AI needs precise garden data. The user doesn't have it.
I designed the system that bridges that gap — an AI-powered site assessment that turns uncertain homeowners into confident gardeners.
Wildlight is an AI-powered web app that helps homeowners plan and maintain sustainable gardens.
I worked as a Product Designer at Accenture, embedded with their team and stakeholders. The product launched internally at Accenture's Season of Impact event and is now being pitched to major retailers including Home Depot.
ROLE
Product Designer
Timeline
Jan-April, 2026
The original approach assumed users already knew their garden details — site size, soil type, sun exposure, drainage.
For most homeowners, that's not the reality.
We moved to an AI-first model — replacing the static form with a guided conversational flow where the system asks one question at a time and adapts based on user responses.
THE APPROACH
Designing the conversation, not just the interface
The site assessment is a guided AI conversation, not a traditional form. I mapped out what the user does and what the LLM does at every step: when the system asks a question, when it offers help, when it educates, and when it moves on. Every screen-level decision traces back to this interaction flow.
USER TESTING
What the research revealed
After redesigning the flow to be AI-first, we tested it with 20 users across varying experience levels. The direction was validated, but three critical friction points surfaced that needed to be resolved.
Finding 1
14 of 20 participants guessed rather than engaged — the intake felt like a form to rush through.
Finding 2
Square footage was meaningless without relatable comparisons.
Finding 3
water drainage, sun exposure, and soil type confused 13 out of 20 participants
KEY DESIGN MOMENTS
Three insights, three design responses
01
Reframing the intake as guided discovery
Users rushed through the original intake because it felt like a form gating access to the plan. 14 of 20 participants guessed on at least one question rather than slowing down to engage with it. The final design reframes the assessment as guided discovery. Onboarding cards communicate why accurate site data matters before users even begin, and the system builds one question at a time so users understand that getting this right is the work.
03
Designing for uncertainty
When users hit a question they can't answer, they can tap "I'm not sure" to surface an educational module — a short video or diagram vetted by Accenture's subject matter experts will appear.
This keeps users in the flow instead of abandoning or guessing. The AI isn't just asking — it's teaching.
02
"I don't know my garden size"
Instead of asking users to estimate square footage, I designed a polygon map tool that lets them draw their garden boundary on a satellite view. The system calculates the area automatically.
This turns a question most users can't answer into an interaction most users can complete in under 30 seconds.
Outcome
Where it landed
The site assessment shipped as part of Wildlight's internal launch at Accenture's Season of Impact event and was showcased at Web Summit Vancouver 2026. The product has generated interest across Accenture's global teams and is being pitched to major retailers including Home Depot.
