Meet HeatChat

Designing an AI You Want To Talk To

I wanted HeatChat to be more than an AI assistant. It needed to feel alive, trustworthy, and human.


Early prototypes were robotic and generic. Users could not relate to them, and the AI lacked the confidence, energy, and insider knowledge that Heat embodies.


The challenge was clear. Design a chat experience that feels like a trusted friend who knows the city, blends class and street awareness, and speaks the Heat language without ever sounding like ChatGPT.


My Role

I led HeatChat end to end, combining UX, copywriting, psychology, and AI prompt strategy:


  • Conceptualized the AI’s persona and tone

  • Designed interaction flows and dynamic UI (chat + map view)

  • Trained the AI to produce human-like, context-aware responses

  • Integrated it into Heat and collaborated with developers

  • Created microcopy, prompts, and messaging frameworks

  • Marketed HeatChat as the go-to insider guide

Process

Research and Community Feedback

I knew assumptions alone were not enough. HeatChat had to reflect real user expectations.


  • We asked our followers and community how they pictured the AI:

  • Male or female

  • Sophisticated or playful

  • Classy or street-smart

  • Synthesized responses to define persona traits that were relatable

Persona and Tone Design

Next, I gave HeatChat a face via MetaHuman and built a personality that embodied class and street awareness.


  • Defined traits: witty, confident, approachable, insider

  • Created a tone map for different interaction types (greeting, recommendations, errors, encouragement)

  • Outlined dos and don’ts to stay on-brand


Visual Placement: MetaHuman image and tone map diagram

Copy and AI Prompt Framework


Microcopy was everything. I built prompt frameworks to guide AI responses:


  • Example prompts for recommendations, greetings, and conversational context

  • Iteratively edited outputs to humanize tone

  • Avoided anything “ChatGPT-like.” The AI sounds like a friend with insider knowledge


Visual Placement: Sample prompts → AI output → humanized HeatChat response

Testing and Iteration


  • Tested prompts and flows with internal users

  • Iterated based on tone, clarity, and engagement

  • Early errors were corrected and became learning examples for the AI’s future behavior


Visual Placement: Screenshots of early mistakes versus final messages

HC Early Mistakes

Results and Impact


  • HeatChat now feels like a trusted, human-like insider and not just a generic assistant

  • Users connect with it as a friend who knows the city, increasing engagement and retention

  • Developed a repeatable framework for AI UX: persona → tone → prompts → microcopy → integration

  • Blended psychology, UX design, and copywriting into a cohesive product that feels personal, fun, and emotionally intelligent



Visual Placement: Video of you explaining HeatChat or real-life chat screenshot

Feature title.

Feature description.

Feature title.

Feature description.