AllGen: A Generative AI Copilot with Customizable APIs and Role-Based User Management for Enterprises


Context
AI Copilot Domain, US 🇺🇸. Timeline: Q3-Q4 2023.

Role
Design Lead, User Research, Lo-fi & Hi-fi Prototyping, User Testing.

Tools
Figma, FigJam, Adobe Illustrator
Overview
Objectives
Approach
AITP offers dedicated generative and “agentic” solutions addressing operational processes across enterprises. These solutions are designed to drive out labor, cost, and inefficiency from key operational processes, with advanced capabilities including model selection and tuning, workflows, and automation.
To ensure generative AI success, AITP provides comprehensive enablement, training, and end-user support. This includes customized training aligned with business objectives, task-specific training for users, managed internal user communities, and support for generative AI tools like Microsoft Copilot, ChatGPT, Claude, and AITP’s EnterpriseGPT.
Â
AITP’s EnterpriseGPT System is a secure, customizable, and affordable “ChatGPT-like” system providing generative AI power to entire companies. It is deployed in clients’ Azure tenants or dedicated Azure environments managed by AITP, ensuring data protection. The system uses the latest LLMs (e.g., GPT-4o) customized with company knowledge and includes comprehensive enablement and support through AITP Advance.

Userflow
Admin (AITP)​
Creates organizations upon product purchase Sends invitations to organization leads Accesses a dashboard with a list of organizations (status: active/expired, join date, etc.)
Organization
Creates organizations upon product purchase Sends invitations to organization leads Accesses a dashboard with a list of organizations (status: active/expired, join date, etc.)
Users (Employees)​
Use the platform’s AI capabilities Create and manage prompts Interact with AI models for day-to-day workflows

Final Product: Key Features

1. Model Selection
2. Customizable Prompt Library


3. API Key Management
Chat Screen

Outcomes
🚀 Launch & Early Adoption
We launched the test version in November 2023 after a 3-month development cycle. To speed up delivery, we reused existing Ant Design UI components. We celebrated the launch over a video call, and the client successfully onboarded initial organizations from their network. Post-launch, we focused on fixing bugs and iterating based on user feedback.
Learning
🎓 Technical Growth & Future Vision
This was my first AI-focused project, and I gained hands-on experience with how model switching and API integrations work. I also learned how to design around these technical requirements. The client’s next goal is to explore features like vector database uploads, image generation, and expanding AI capabilities on the platform.