Progressive Loading
Meeting metadata can appear while slower transcript and AI work continues in the background.
AI SaaS / Meeting Intelligence
ElevateAI
ElevateAI was built for people who want the usefulness of a meeting assistant without manually organizing every note afterward. The product lets a user create a domain-specific AI agent, meet with it in a live video session, and leave with transcripts, summaries, recordings, and a chat surface that keeps the meeting useful after the call ends.
Type
AI SaaS / Meeting Intelligence
Timeline
Built as an end-to-end product sprint, from prototype to deployed SaaS experience.
Focus
Product engineering
Overview
The motivation was to explore what an AI-native meeting workflow should feel like when the assistant is part of the product from the beginning. Instead of treating transcription, summarization, and follow-up chat as separate tools, ElevateAI connects them into one loop: prepare an agent, run a meeting, capture the conversation, then ask questions against that meeting context.
The product is aimed at builders, students, creators, and small teams that need structured conversations but do not want heavyweight meeting software. The experience stays direct: choose an agent, start a session, and receive useful artifacts with minimal ceremony.
Technical Details
- Next.js: Used for the product shell, route-based rendering, API boundaries, and deployment-friendly performance.
- TypeScript: Keeps agent, meeting, transcript, and summary data predictable across screens and server calls.
- Tailwind CSS: Supports fast iteration on dense product UI such as meeting controls, panels, and empty states.
- AI APIs: Transform raw conversation into summaries and contextual follow-up answers.
- Vercel: Provides low-friction hosting, preview deployments, and environment configuration.
Data Flow
01
A user creates or selects an AI agent with a clear meeting role.
02
The meeting session starts and produces transcript and recording artifacts.
03
Transcript content is processed into a structured summary after the meeting.
04
The chat interface receives the meeting context and answers follow-up questions from that source material.
Features & Functionality
- Custom AI Agents: Users can create assistants for specific meeting contexts.
- Live Video Meetings: The meeting room keeps the conversation inside the same product that will process the outcome.
- Automatic Transcripts: Meetings produce text artifacts that can be searched, summarized, and reused.
- AI Summaries: After the meeting, the product produces a concise summary of the important points.
- Post-Meeting Chat: Users can continue asking questions about a completed meeting.
Performance Optimizations
Route-Level Splitting
Next.js keeps setup, meeting, and review screens separated so users do not load every interface at once.
Image Reliability
Static and remote images are configured to avoid broken production renders.
Loading States
AI workflows are not instant, so explicit states reduce uncertainty during processing.
Development Process
- 1
The project began with the question: what should happen after a meeting ends if the AI assistant actually understands the meeting?
- 2
The first prototype focused on starting a session, capturing useful text, and producing a meaningful summary.
- 3
The UI was then refactored around the journey: agent setup, live session, generated artifacts, and follow-up chat.
- 4
Final polish centered on deployment behavior, image reliability, and making the post-meeting experience feel continuous.
Challenges
Coordinating Real-Time and Async Work
A meeting happens live, while transcript processing and summaries happen after the fact. Each stage needed to be explicit in the product flow.
Keeping AI Output Useful
Summaries need structure without sounding generic. Focused context and role-specific agents guide generation.
Avoiding UI Overload
Meeting tools can become crowded quickly, so controls and generated content are separated by stage.
Deployment
- Vercel Hosting: Vercel keeps Next.js deployment, previews, and environment configuration straightforward.
Future Improvements
Team Workspaces
Shared agents and meeting history would support recurring team workflows.
Action Item Extraction
Turning summaries into structured tasks would reduce the gap between conversation and execution.
Search Across Meetings
Cross-meeting search would help users retrieve decisions without remembering the exact session.
Lessons Learned
- AI Products Need Workflow Design: The hard part is not only calling a model; it is deciding when AI output appears and how users act on it.
- Async UX Matters: Meeting artifacts arrive at different times, so loading and empty states are part of the product architecture.
- Context Is Product Value: The post-meeting chat becomes useful only when it is grounded in the specific session.
Final Thoughts
ElevateAI is a strong example of building around a complete user loop instead of a single AI feature. It turns conversation into durable knowledge and makes the next interaction with that knowledge feel natural.