Work
Convay AI for Physical Meetings
Convay AI for Physical Meetings started with a simple question: Why should only online meetings be easy to document? In high-stakes rooms, whether it's a government briefing or an internal sync, people still scramble to take notes or forget key points entirely. This feature was our answer: record audio, let AI handle the rest. Transcripts, summaries, multilingual support, all packaged into a simple, intuitive flow that turns real-world conversations into shareable outcomes.
- Role
- UX Designer, UI Designer, UX Researcher
- Category
- Web Design, Product Design
- Duration
- Sep 2023 – Jul 2024 (Convay tenure)
- Tools
- Figma
A first look at Convay AI for Physical Meetings, from the meeting dashboard to a finished, shareable summary.
Convay at a Glance
Convay is a video conferencing and collaboration platform built for governments, enterprises, and institutions that need to keep control of their own data. Used in more than 45 countries, it covers a meeting from start to finish, from scheduling and live video to AI transcription, file storage, and follow-up, all inside a system that can run entirely on a client's own servers.
Convay is the platform behind Bangladesh's national government communications system, built for the Bangladesh Computer Council and designed to scale to 200,000 users. In 2025, it was also selected over Zoom and Google Meet as the official platform for the Sustainable Infrastructure Development Symposium (SIDSSA), a G20-linked summit hosted in South Africa, supporting government officials and policymakers from 27 countries.
At Evercare Hospital Dhaka, a 425-bed hospital and the first in Bangladesh to receive JCI accreditation, Convay has run telehealth appointments and meetings for two and a half years, according to the hospital's head of IT.
- Countries
- 45+
- Government platform capacity
- 200,000
- Participants supported in a single meeting
- 10,000
Capabilities
- Enterprise-grade security: sovereign deployment on cloud, on-premises, or hybrid, with data kept in the client's jurisdiction
- High-quality video and audio: built for low-bandwidth and unstable network conditions
- AI transcription and meeting summaries: around 93% accuracy, in five languages: English, Bangla, French, Portuguese, Spanish
- Real-time whiteboard, chat, and screen sharing: with cloud storage for meeting files and chat logs
- Cross-platform: web, iOS, and Android, with 10,000+ mobile installs
References
Explore ConvayThe Problem
In high-stakes in-person meetings, like government briefings or internal syncs, crucial details often go undocumented. Teams either take rough notes or rely on memory, risking missed tasks, misunderstood decisions, and accountability gaps. Traditional transcription tools cater mostly to online settings, leaving physical conversations in a productivity blind spot.
Convay needed a way to bring offline meeting intelligence into its digital ecosystem, without making users change their natural workflow. The goal was clear: create a feature that could automatically capture, structure, and summarize real-world discussions with the same quality and ease as online calls. It had to be simple enough for a 10-minute sync, yet secure and reliable for official proceedings.
Online meetings
Recorded, transcribed, and summarized automatically.
Physical meetings
Handwritten notes, memory, or nothing at all.
The gap this feature closes.
My Role and What I Built
I was part of the core design team during the early concept and prototyping phase of this feature. My focus was to translate a vague product idea into a clear, usable experience.
I designed the full user flow, starting from how someone would set up an in-person meeting, to how they'd record, review, and extract insights after the session. I created UI mockups, built multiple design variants for testing, and worked closely with developers to ensure technical feasibility. My decisions were guided by real-world use cases, like small internal syncs and government meetings where formal documentation was essential.
I also ran internal usability testing to compare interface versions, gathered feedback, and iterated based on what worked best for the users. Every design choice aimed to reduce friction and make the tool feel as seamless as pressing "Record" and getting a polished summary back.
What I Built
Convay AI for Physical Meetings brings offline conversations into the digital workspace. Instead of relying on handwritten notes or memory, users can now record in-person meetings, generate full transcripts, summarize them with AI, and export everything, seamlessly. The feature was designed to be fast, accurate, and secure, especially for confidential internal syncs and government use cases. Whether it's a small strategy meeting or a closed-door briefing, teams leave with clear, shareable documentation, just like they would in a virtual meeting.
This feature also supports multilingual transcription, including English and Bengali, making Convay one of the first platforms to support real-time, localized summaries for physical meetings.
Design Process
Research
We began by identifying a key gap in physical meetings: no easy way to document discussions in real-time. Through internal interviews and team feedback, we confirmed that manual notes often led to missed details, especially in fast-paced syncs or confidential meetings. This insight guided the idea: make in-person meetings as easy to document and summarize as online ones.
Ideation and Design
We mapped a simple, linear flow. Each screen was designed to be clean, minimal, and mobile-friendly to support quick use in real-world meeting environments.
- 1
Set meeting context
- 2
Record audio
- 3
Generate transcript
- 4
Summarize
- 5
Export
Every screen in that flow is walked through in the next section.
Iteration and Testing
During testing, we developed two versions of the Meeting Panel page to evaluate user preferences.
A / B
We also tested the transcript editing and summary flow, ensuring that each step felt intuitive and could work without onboarding.
Feedback Led To
- Removing the "Mute" text label (icon only)
- Refining the AI summary button placement
- Smoothing transitions between transcript → summary → export
The Flow, Screen by Screen
The complete flow, grouped by phase, from setting up a meeting to walking away with a finished summary.
Setup
Starting a session takes two screens: find the meeting, then name it.
Meeting Dashboard: access all meetings, start new ones, upload a recording, or view recent recordings.
Meeting Information: enter a title and description, then Start or Cancel.
Record
The recording screen itself, and the one confirmation step that protects against ending a session by accident.
Main Meeting Panel: a large microphone icon and live transcript display.
End Session Confirmation: prompts users to confirm ending the session.
Process
Once a session ends, transcription and summarization run automatically, each with its own loading state.
Loading UI for Transcription: a visual indication of the transcription process.
Meeting Transcription: view and edit the transcription.
AI Summarization Loading: a loading skeleton while the AI summarizes the meeting.
Meeting Summary: displays the summarized content.
Export & Edit
The last stretch covers getting the output out the door, and fixing anything the transcript got wrong along the way.
Export Options: downloading the meeting summary and transcript.
Transcript Editing Options: a 3-dot modal for editing the transcript and renaming speakers.
Edit Transcript: make changes to the transcript.
Rename Speaker: renames a speaker throughout the transcript.
Edit Meeting Title: modify the meeting title.
Edit Meeting Title: a new title being entered.
Edit Meeting Title: the updated title after confirming.
Challenges and Solutions
Noisy physical environments reduced transcription quality. In real-world settings like government offices and team syncs, background noise made it difficult to capture accurate transcripts.
We explored early-stage noise filtering and speaker identification techniques during research. While still in development, internal tests showed a notable boost in accuracy, especially in quieter rooms.
Even in busy rooms, the transcript quality held up well.
Inconsistent UI between Meeting Panel and core platform. The Meeting Panel was originally designed in light mode, while Convay's main UI used dark mode, causing visual dissonance and distraction.
We tested both variants (shown side by side above) and switched to dark mode for consistency and better focus.
→ 82% of internal testers preferred the updated dark version.
Speaker renaming was time-consuming. Users needed to manually update speaker names throughout the transcript, especially frustrating in longer meetings.
We designed a bulk rename flow, allowing users to rename once and apply it throughout.
→ This reduced speaker editing time by an estimated 30-40%.
Users were unsure about using AI summarization. Some users hesitated to click "Summarize with AI," unsure of its purpose or outcome.
We gave the button a clear label, prominent placement, and optional use, ensuring it felt helpful, not overwhelming.
I didn't expect it to be that simple. The summary saved me a ton of time.
Outcome and Impact
Though still in its early stage, Convay AI for Physical Meetings delivered strong results in internal testing and team evaluations.
- 93% transcription accuracy was achieved in quiet environments, even with varied speaker tones, based on internal testing rather than a large-scale field study.
- Teams reported a 30-40% reduction in time spent creating meeting notes, an internal estimate from the same testing.
- The AI summarization feature helped users quickly recall key takeaways, especially in confidential internal syncs where formal minutes were previously missing.
- Support for Bengali transcription gave Convay a unique edge over global competitors, especially for local government adoption.
- The streamlined export process (PDF) made it easy to share documentation post-meeting, helping teams align faster.
This feature also added a strategic angle: it positioned Convay not just as an online collaboration tool, but as a Zoom + Otter.ai hybrid, extending value even in offline and hybrid use cases. As Convay scales to 10,000+ participant support and continues to serve government ministries and international symposiums, this feature shows how UX can drive product evolution and market differentiation.
Takeaways
Designing Convay AI for Physical Meetings helped me think beyond screens and solve for real-world complexity, like making offline meetings as actionable as online ones. I learned to simplify AI workflows, design multilingual support with accessibility in mind, and ensure consistency across light and dark modes.
This project also sharpened my collaboration skills with developers, especially while testing transcript editing logic and speaker detection flows. Most importantly, it reminded me that great UX isn't just about clarity, it's about trust. Especially when the users are government teams, confidential briefings, or high-stakes internal syncs.
I now approach every feature with the mindset of:
How can this reduce friction and make someone's job easier, even when they're not behind a screen?