Live assistance is only useful if it understands the right conversation. Showing a generic suggestion in another window isn't enough: it needs to know which prospect is speaking, which offer is being presented, what goal was prepared, and what happened a few minutes earlier.
The Chrome extension bridges the video meeting and the Objeqo workspace. It doesn't recreate the call. It finds the one that's already been prepared, captures authorized audio sources, sends usable segments, and displays only the information that's useful in the moment.
One call, from preparation to report
- 01
Prepare
The sales rep selects an offer, enters the prospect's details, and generates a brief with goals, questions, and anticipated objections.
- 02
Join
From the extension, they find that same call on Google Meet, Zoom, or Microsoft Teams, without creating a parallel session.
- 03
Capture
The microphone and tab audio are used after an explicit action and with browser permissions.
- 04
Analyze
The transcript progresses in segments. Memory, signals, and objections are updated on the server.
- 05
Report back
Suggestions stay brief during the call, then key moments inform the report and coaching.
Why the microphone and tab are two separate sources
The microphone primarily captures the sales rep's voice. The tab captures the other participants and meeting sounds. Using just one source would produce an incomplete transcript, with answers but no questions or objections that are impossible to attribute.
The extension coordinates these streams when the sales rep explicitly starts the call. It shows a clear status for the permission request, start, activity, and stop. Capture should never seem implicit.
The transcript becomes a living memory
Waiting until the meeting ends to produce a long text would make live assistance impossible. So the conversation comes in segments that preserve its chronology. Each new passage adds to what the system already knows about the call.
A segment on its own can be ambiguous. Before analysis, it is matched with the offer, prospect, brief, goal, and recent moments. This continuity makes it possible to distinguish a simple question from a likely objection and suggest a response that takes into account what has already been said.
| Item | Role during the call |
|---|---|
| Brief | Recalls the stakes, questions, arguments, and anticipated objections. |
| Recent segments | Provide the immediate context for the phrase heard. |
| Memory | Retains facts, constraints, and decisions already expressed. |
| Signals and objections | Indicate points of progress to verify or explore further. |
| Call ID | Keeps all data associated with the right conversation. |
- Item
- Brief
- Role during the call
- Recalls the stakes, questions, arguments, and anticipated objections.
- Item
- Recent segments
- Role during the call
- Provide the immediate context for the phrase heard.
- Item
- Memory
- Role during the call
- Retains facts, constraints, and decisions already expressed.
- Item
- Signals and objections
- Role during the call
- Indicate points of progress to verify or explore further.
- Item
- Call ID
- Role during the call
- Keeps all data associated with the right conversation.
Sensitive analysis stays on the server
A browser extension can be inspected. So it should never contain a private key used to call a model. Authenticated routes receive only the necessary context, verify the user, and perform the analysis in the server environment.
Security also depends on data isolation. Each brief, segment, detection, and report must stay associated with the right user and call. Context from another conversation must never be used to generate a suggestion.
- No private OpenAI key in the extension code.
- Authenticated server routes for analysis.
- Access control for call data.
- Browser permissions requested when needed.
- Secrets kept separate from explicitly public values.
The call must continue even if assistance slows down
The network can slow down, a segment can arrive late, and an analysis can fail. These issues affect the assistance, not the conversation itself. The extension must therefore separate capture, transcription, and suggestion generation.
| Incident | Expected behavior |
|---|---|
| Suggestion unavailable | The call continues, and another attempt can be offered. |
| Delayed transcript | The status is visible without blocking the call controls. |
| Audio source missing | The user can see which permission or source needs to be corrected. |
| Session expired | Reconnecting is required without mixing data with another session. |
- Incident
- Suggestion unavailable
- Expected behavior
- The call continues, and another attempt can be offered.
- Incident
- Delayed transcript
- Expected behavior
- The status is visible without blocking the call controls.
- Incident
- Audio source missing
- Expected behavior
- The user can see which permission or source needs to be corrected.
- Incident
- Session expired
- Expected behavior
- Reconnecting is required without mixing data with another session.
The guiding principle
Connect the context without complicating the call.
The extension works as a continuity layer. It connects preparation, conversation, and reporting without replacing the video conferencing platform or asking the sales rep to rebuild their context.
This unobtrusive experience depends on a rigorous architecture: two audio sources, a progressive transcript, server-side analysis, strict data isolation, and errors that never take control of the call.
Explore further
Continue with a topic related to this conversation.
Frequently asked questions
Frequently asked questions about the Chrome extension
Which platforms can I use the extension with?
The workflow is designed for Google Meet, Zoom, and Microsoft Teams in Chrome, with microphone and tab audio capture when permissions are granted.
Do I need to recreate the brief from the extension?
No. The extension reuses the call and brief already on file. Preparation stays linked to the same conversation when you schedule, join, or restart it.
Is the OpenAI key included in the extension?
No. Model calls are made on the server. A private key must never be exposed in the browser or in the extension package.
What happens if a suggestion fails?
The call continues. The interface can report the error and offer another attempt without blocking capture, the conversation, or preparation features.