PromptMesh –
One Interface, Multiple AI Models
Background & Problem Statement
PromptMesh is a productivity-focused AI tool that enables users to access and compare responses from multiple large language models (LLMs) such as GPT-4, Claude, and Gemini in one unified workspace. The product was designed to reduce context-switching, improve AI response quality, and help users discover which model works best for their specific workflow.
AI tools are rapidly evolving, but users often find themselves jumping between multiple chat platforms to compare model performance. Developers and researchers want fast, side-by-side evaluations without duplicate typing or copy-pasting. PromptMesh aimed to solve:
Too much switching between different AI chat apps
No easy way to compare output quality across models
No chat memory or history organization between sessions
Research & Discovery
Methods:
8 stakeholder interviews across product, HR, finance, and engineering teams.
Competitive analysis of: ChatGPT, Gemini, Claude, Notion AI, Grok, Slack AI, ChatGPT Teams, Writer.
Internal surveys to understand where AI is used and where it fails in collaboration.
Insights:
Prompt reuse is low despite repetition across teams.
Admins want audit logs, usage analytics, and role-based permissions.
Users prefer structured templates with clear variables.
Most current tools focus on chat, not knowledge sharing or AI governance.
User Research & Synthesis
I conducted interviews with 6 developers and researchers who frequently use ChatGPT, Claude, and Gemini. We mapped their needs into clusters using sticky note synthesis.
Color-coded quotes highlighting frustration with testing flows, output quality, and chat history management.
Top insights:
“I want to type once and see every model’s answer together.”
“I don’t care which model is better globally, I care which works best for my domain.”
“I need a way to track which models gave me useful outputs and save them easily.”
Borrowed Wisdom
Taking inspiration from the already established tools
We didn’t start from scratch—we studied what was working elsewhere. ChatGPT provided clarity in our chat history. Gemini showed us the power of inline context. Claude was a master of summaries. Grok added tone and speed. Notion AI reminded us how format matters.
But none of them focused on teams.
PromptMesh took the best of each and layered it into a system designed for multiple minds working together—structured, collaborative, and reusable.
We analyzed core UX strengths and limitations of existing tools to guide our design strategy:
📎 ChatGPT: Clear conversation history and prompt chaining, but lacks workspace-level collaboration
📚 Gemini: Structured UI with inline references, inspiring how we handle template annotations
🤖 Grok (X): Tone customization and snappy UX guided our casual tone support within templates
🧠 Notion AI: Document-based prompting influenced our "prompt-to-template" save flow
🧩 Claude: Useful for long-form generation and summaries — we tested prompt styles that maximize results across tools
None offered direct, real-time model comparison or prompt syncing.
From Idea to Interface
We sketched before we shipped. Early concepts explored what a collaborative prompt builder might look like, how to toggle between LLMs, and what a library of shareable prompts should feel like.
It wasn’t just about features—it was about reducing friction and building trust. We prioritized clarity, discoverability, and speed—so people could move from idea to execution without switching tools or losing flow.
Low-fidelity sketches for:
Prompt Builder modal (with variable fields)
Multi-model chat (tabs and vertical split views)
Template browsing grid
Focused on:
Reducing cognitive overload
Making chat + content generation collaborative
Simplifying access to shared resources
Design Gets Real
This is where vision met visuals. With structure in place, we brought PromptMesh to life—every screen designed to guide, not overwhelm.
We used color to communicate trust, motion to suggest intent, and layout to remove friction. The result? A modular, expressive interface that makes collaboration feel effortless.
Key Screens:
AI Chat (Compare GPT-4, Claude, Gemini)
Template Builder (structured inputs + outputs)
Admin Analytics (filters, activity stream, export)
Results and Metrics
We conducted a beta test with 40 users over a focused 2-week period to evaluate performance, usability, and value across real workflows:
92% of users found PromptMesh useful for streamlining multi-model workflows, while 8% needed more guidance.
Most users tested multiple LLMs within a single conversation—reducing platform switching.
20+ custom prompt templates were created and shared during testing.
What I’ve Learned
I’ve learned that the true power of AI lies in collaboration—not just smarter assistants, but systems that help teams reuse knowledge and move faster together.
Right now, I’m focused on launch prep: refining documentation and building Workflow Agents that turn repetitive tasks into automated flows. It’s early, but I can already see where PromptMesh is going—and it’s exciting.