AI Solar Advisor: Helping users make complex solar decisions with confidence

Summary

Based on user research, we created a first-of-its-kind tool directly addressing a major pain point blocking people from switching to solar.

Context

The company

WattBuy, Series A clean energy startup helping consumers switch to cleaner energy through data-driven recommendations.

My role

Founding designer

Product

(B2)B2C energy platform – consumer marketplace for electricity and solar, plus APIs for enterprise partners.

This project

A data and AI powered chatbot that helps users get answers when they don't know what questions to ask and who to trust.

What I did

Designed the chatbot character, interface, and prompt.

Other players

Collaborated with CPO and an engineer to prioritize data sources. Engineer implemented data ingestion and RAG system.

Timeline

~2 months

Problem definition

Why an AI chatbot, why now

Business context

Our newly developed PMF stated that we were "a one-stop shop for solar, empowering homeowners with trustworthy, personalized, actionable information and tools to enable their switch to solar energy."

To that end, we were developing a suite of B2B2C solar tools as offerings in addition to our APIs.

I ran user research that showed

The biggest pain points existed in the initial research phase, where homeowners have trouble due to:

Complex, scattered information not relevant to their home
Personalized details locked behind pushy sales conversations
No trusted, neutral source for personalized guidance

The landscape

Existing tools: rigid calculators that require specific house knowledge; conversations with pushy installers
Gap: No accessible, personalized, and trustworthy guidance for a complex, high-stakes decision
Timing: Maturity of GenAI tools (GPT-4, embeddings, RAG) + proprietary energy data = first-mover advantage.

Solution

Sunny Wattkins, an AI-powered solar advisor

Leveraging WattBuy's APIs for personalized usage and cost data and ChatGPT's conversational abilities, Sunny provides accessible solar guidance that helps curious homeowners explore their options independently.

1Character development

User research revealed people trust the "friendly neighbor with solar experience" archetype, which became the inspiration for Sunny's personality.

2Starter prompts

I identified common solar questions through user research and installer feedback to provide immediate value.

3Personalized data

We used WattBuy's APIs to provide personalized data, including usage and cost estimates, to help users make informed decisions.

4Feedback

By reviewing conversation and user feedback, I refined the prompt to make Sunny's responses clearer, faster, and more useful.

Challenge

From interrogation to conversation

Problem

Sunny aggressively demanded address for any info. Unproven value + sensitive ask = user drop-offs
Instructions getting lost in too long prompt, resulting in less impactful answers

Solution

Redesigned prompt strategy: education before conversion to build trust by demonstrating value
Streamlined prompt length, focusing on essential instructions at start and end

Outcome

Analyzed 1,000+ conversations showing 3 key behavioral improvements:

More questions answered per conversation
Reduced repetitive queries as users got quality answers faster
Fewer clarification requests needed

User feedback: 21 👍 vs. 1 👎 (95% positive)

Before

After

Outcome

AI with real utility

As an R&D exploration for a new user base, we focused on qualitative validation metrics: user engagement patterns, question specificity, and concept desirability. Results:

Answers to specific, detailed questions not found on Google.
Shorter, less repetitive chats, with users leaving after getting needed estimates rather than from frustration.
A new B2B2C partner (a popular local community platform).

This is a very helpful tool and all in one space to discuss. Much appreciated. Sunny is great and it would be nice to get a full report that can be printed out. Smart move to be a part of [partner name].”

— Submitted through Sunny's feedback form