Build a Voice AI Agent That Joins Google Meet and Zoom

Here is a walkthrough of how to build and deploy a meeting bot:

What if you could send an AI agent into a meeting, and it just handled things?

Not a bot that records and transcribes. A voice AI agent that actually talks: conducts interviews, coaches your sales reps, runs screening calls, or silently observes and scores performance against a rubric you define.

That’s what Tough Tongue AI does. It’s a no-code platform for building voice AI agents. You describe what you want the agent to do in plain language, and it handles the rest: real-time voice conversation, multimodal understanding (tone, facial expressions, body language), and post-call analysis.

The part that changes the game: these agents join your actual Google Meet and Zoom calls. Open meetings, private meetings, authenticated rooms. It doesn’t matter. Your agent shows up, does its job, and sends you the results.

This post walks through how to build one from scratch.

---

What You Can Build

Tough Tongue AI supports two types of meeting agents:

Streaming agents show a video avatar and actively participate in the conversation. They talk, listen, ask follow-up questions, and use tools like whiteboards, slides, and image generation. Think of these as autonomous participants.

Notetaker agents join silently. They don’t speak, don’t show video. They observe, record, and then evaluate the conversation against a scoring rubric you define. You get a detailed report after the call.

Here are some concrete use cases:

• First-round screening interviews. The agent conducts the interview autonomously, asks your questions, probes on follow-ups, and scores the candidate. You review the analysis afterward.
• Sales call coaching. A silent notetaker joins your discovery calls or product demos. It evaluates opening hooks, personalization, feature-to-benefit translation, prospect engagement, objection handling, and closing strength. Each dimension gets a 1-10 score with specific quotes as evidence.
• Investor pitch observer. Sits in on your fundraising meetings and scores narrative clarity, financial fluency, team credibility, and closing momentum.
• Training and roleplay. An interactive coach that runs practice sessions with your team. Rehearse difficult conversations, negotiations, or client calls with an agent that adapts in real time.

All of these are defined through scenarios: plain-language descriptions of what the agent should do, how it should behave, and how it should evaluate performance.

---

Step-by-Step: Creating Your First Meeting Agent

Step 1: Create a Scenario

Go to the Scenario Library and click Create New Scenario.

A scenario defines three things:

1. Role. Who is the AI agent? (“You are a senior PM interviewer conducting a product sense round” or “You are a silent note-taker for sales discovery calls”)
2. Agenda. What should the agent do during the conversation? (List of questions, coaching flow, observation criteria)
3. Evaluation rubric. How should performance be scored afterward? (Dimensions, weights, scoring bands)

You write all of this in natural language. No code, no configuration files. Here’s a minimal example:

You are a sales discovery call coach. Stay silent throughout the call.
Observe the rep's performance on:
- Opening and agenda setting
- Discovery depth
- Talk-to-listen ratio
- Closing and next steps

Score each dimension 1-10. After the call, provide the rep's top strength,
biggest missed opportunity, and one concrete drill to practice.

For notetaker scenarios, toggle on Notetaker Mode and Audit Mode in the meeting configuration. This tells the agent to stay silent and focus on observation.

Step 2: Access the Meeting Bot

Navigate to: Library > your scenario > Meeting Bot tab.

This is the control panel for deploying your agent into meetings.

Step 3: Deploy to a Meeting

You have two deployment modes:

Custom Meeting Mode, for one-off or ad-hoc meetings:

1. Select your platform (Google Meet or Zoom)
2. Paste an existing meeting URL, or select “Create New” to generate a fresh Google Meet room
3. Optionally schedule the bot to join at a future time
4. Hit deploy. The agent joins the call

Calendar Integration Mode, for hands-free automation:

1. Connect your Google Calendar (read-only access to detect your meetings)
2. Set keyword filters like “interview”, “discovery”, or “coaching”
3. The agent automatically joins meetings whose titles match your keywords

With calendar integration, you can have different agents handling different types of meetings. A screening agent joins anything with “interview” in the title. A sales coach joins “discovery call” meetings. Each scenario has its own calendar connection and keyword filters.

---

Open vs. Private Meetings: The Authentication Story

This is where most meeting bots fall short. They work fine for open meetings but can’t get into anything with restricted access. Tough Tongue AI’s agents join as authenticated, signed-in Google accounts, not anonymous link-clickers.

Here’s how it works depending on your meeting’s access settings:

Open Meetings

If your Google Meet is set to “anyone with the link can join,” the agent walks right in. Zero friction, no extra steps.

Private / Protected Meetings

For meetings where only invited participants can join, you have two options:

Option A: Manual admit. The agent knocks on the meeting door. A participant sees a “someone wants to join” notification in Google Meet and clicks “Admit.” This works well for notetaker scenarios where someone in the meeting expects the bot.

Option B: Auto-join via email invite. Add ttai@toughtalkai.com as an attendee on the calendar event. The system recognizes the bot as an invited participant, and it joins without knocking. No human intervention needed.

Which approach should you use?

Use case	Recommended approach
Notetaker / passive recording	Open meeting or manual admit
Someone in the meeting can let the bot in	Manual admit
Agent-led coaching, screening, or interviews	Add the email for auto-join
Fully autonomous operation (no humans present initially)	Add the email for auto-join

The auto-join approach is critical for autonomous agent scenarios. If your agent is conducting a first-round interview and the candidate is the only human in the room, there’s nobody to click “Admit.” Adding the email to the calendar invite solves this cleanly.

For Zoom meetings, paste the Zoom URL directly in the Custom Meeting form. The agent joins the call using the provided link.

---

Bonus: Embed the Meeting Bot on Any Website

If you want users on your website to be able to start a meeting with your AI agent, Tough Tongue AI provides an embeddable widget. One <script> tag, two lines of configuration:

<script
  src="https://app.toughtongueai.com/widget/meeting-bot.js"
  data-scenario-id="your-scenario-id"
  data-mode="button"
  data-title="Talk to AI"
  data-button-text="Start Meeting">
</script>

This renders a floating button (or a bottom strip, with data-mode="strip") on your page. When a visitor clicks it, a meeting is created, the bot joins, and the visitor is sent to the Google Meet room in a new tab.

Under the hood, the widget uses a Shadow DOM shell and an iframe on Tough Tongue AI’s domain, so there’s no CORS configuration needed on your end. It just works.

---

What Happens After the Call

Once the meeting ends, the platform processes the recording:

1. Transcription. Full transcript of the conversation, automatically generated.
2. AI analysis. The agent evaluates the session against the rubric you defined in the scenario. Each dimension gets a score, evidence (specific quotes and moments), and coaching recommendations.
3. Email report. Scores and analysis are emailed to the scenario admin automatically.
4. Session dashboard. All sessions are accessible in the Sessions page with recordings, transcripts, and evaluations.

For teams that need data in their own systems, Tough Tongue AI offers an API. You can pull session results programmatically, pipe them into Google Sheets with the provided Apps Script integration, or build custom dashboards.

The evaluation output isn’t generic “good job” feedback. For a sales discovery call, you’d get something like:

Top Strength: At 4:32, the rep asked “what does that cost you in engineering hours?” This 3-layer-deep question turned a surface-level complaint into a quantified $180K/year pain point.

Biggest Missed Opportunity: When the prospect said “we already have something for that,” the rep moved to the next feature instead of asking “what specifically does your current solution handle well?”

One Concrete Drill: Before your next call, practice the acknowledge-clarify-address framework. When you hear resistance, say: “That’s a fair point. Can you tell me more about what’s working with your current approach?”

---

Getting Started

1. Sign up at app.toughtongueai.com
2. Create a scenario describing your agent’s role, agenda, and evaluation criteria
3. Deploy to a meeting via the Meeting Bot tab. Paste a URL or connect your calendar
4. Review results in the session dashboard or your inbox

The meeting bot is available on Premium ($20/month) and Business ($99+/month) plans. You use your existing platform minutes with no separate metering.

Links:

• Platform
• Google Meet Agent Guide
• Video Demo
• Contact | Book a Call