EnergyBot GPT

Designing the EnergyBot GPT: Translating complex energy shopping into a high-intent, zero-friction conversational flow on the OpenAI platform.

MY ROLE

UI Design

Conversational UX

User Research

Product Strategy

DELIVERABLES

EnergyBot GPT - a custom application on the OpenAI/ChatGPT platform.

TEAM

Me (product design)

Product Manager

Developers (team of 5)

OVERVIEW

The challenge

Translating a complex, multi-step electricity plan comparison e-commerce experience into a minimal-friction, conversational flow compliant with the OpenAI Design Guidelines.

The goal

Achieve the shift from "search and browse" to "chat and decide," leading to higher-intent conversions to the main EnergyBot website.

The opportunity

Capture high-intent users (homeowners and renters) who are already comfortable using AI for decision-making.

SOLUTION

Key interaction flow: progressive personalization

The design uses a Progressive Personalization model to deliver immediate value while minimizing user effort:

1. Low-Friction Start: The app begins by asking only for the user's zip code.

2. Default Value Delivery: It immediately returns a Refined Plan List (maximum of 3–5) calculated using a transparent default assumption (e.g., based on an average usage of 1000 kWh).

3. Refinement Gate: The GPT prompts the user: "These plans are based on 1000 kWh usage. Would you like a more personalized result? If so, just tell me your average monthly bill or usage!"

4. Deep Personalization: If the user provides their exact usage, the GPT recalculates, greatly increasing the likelihood of high-value conversion.

5. Q&A Integration: Users can instantly ask follow-up questions about specific plans (e.g., "What are the early termination fees for Plan A?").

Design process: overcoming Open AI platform constraints

The primary challenge was translating complex data into simple conversational outputs while adhering to the OpenAI Design Guidelines, which mandate minimal text and action.

‍Constraint 1: The Principle of Minimal Information
‍The most challenging design constraint was the platform's encouragement to reduce the information payload.
Design Solution: Focused data cards & progressive disclosure. I prioritized the three most important factors (Price, Term Length, and Overall Cost) based on prior user research. The Progressive Personalization flow minimizes the upfront information collection, adhering to the minimal friction mandate. The key is to display the honest "All-in Rate" (EnergyBot's value proposition) while deferring full legal details to the final website click-through.

‍Constraint 2: Conversational Feedback & Error Handling
‍With the absence of visual UI elements, conversational design (UX writing) was critical for successful interaction.

Design Solution: conversational clarification loops. Instead of simple error messages, the GPT initiates a graceful clarification loop. For example: User: "I want a cheap plan." GPT: "I can certainly find you a cheap plan! Please provide your Zip Code first so I can pull the correct rates in your area." This approach ensures the necessary data is collected for accurate results without breaking the conversational flow.

Outcomes & future impact

The design is engineered to leverage the high-intent environment of the ChatGPT platform to drive transactions on the EnergyBot website.

Key Metrics for Success
‍Our focus is on full-funnel measurement, validating both the design quality and the successful conversion to a completed order:

Conversion rate to website link (primary funnel metric): Measures the effectiveness of the conversational UX in generating a highly qualified, high-intent lead who clicks through from ChatGPT.

Full-funnel completion rate (ultimate success metric): Measures the percentage of users who start the conversational flow in ChatGPT and complete the ultimate goal: successfully placing an order on the EnergyBot website and receiving confirmation.