GoTranscript
>
All Services
>

Public/build A Taco Bell Voice Agent In 5 Minutes

Build a Taco Bell Voice Agent in 5 Minutes (Full Transcript)

A quick demo of Assembly.ai’s Voice Agents API: copy a prompt, answer setup questions, auto-generate code, and run a natural-sounding ordering agent in minutes.
Download Transcript (DOCX)
Speakers
add Add new speaker

[00:00:00] Speaker 1: Hey, Ryan here from Assembly.ai. Wanted to show you how easy it is to get started with our new Voice Agents API. I just signed up for a fresh account with a fake email, and here I am on our activation page. You can actually click here and copy prompt and directly take this prompt into your favorite coding agent and get started with the API. Let's see how it works in action. Now that I have my prompt copied, let's go ahead and paste that into Cloud Code so we can get started with Assembly.ai. This prompt will automatically start prompting us what we want to build. Voice agent for Taco Bell ordering, starting fresh. And so it says, what are you building? And are you adding Assembly.ai to an existing project or starting fresh? We want a managed voice agent through the new Voice Agent API. It's now asking me questions about orchestration, how I want to set things up, et cetera. Let's go ahead and use a web browser for where the audio will come from. But of course you could also use a phone number and something like Twilio or Telnyx. And then it's also asking me which language. You can choose this based on your environment, but I'll go ahead and say, you know what, let's have fun. You pick, it's just a demo. And so you can kind of see very quickly, like what's your use case, how we want to set this up, et cetera. And I have a key. Let's go ahead and paste the API key that we had from the signup page into the session here. Of note, security wise, would definitely not recommend recording your API keys. This is entirely a demo and I'll be rotating this afterwards. And after that, US or EU, US only. So it's asking us questions about where we want to actually send the data, you know, to the US data center, the EU data center, et cetera. And it's going to use all of this input to actually automatically go and generate all of the code that we need to have a voice agent with Assembly. So now that it's at this point, it's actually talking you through what it's planned. It's a browser-based Taco Bell ordering voice agent. It's going to use the voice agent API from Assembly API. It's going to be in the US region and you can see a lot of other key information about that on the screen. It's going to start building it now from scratch. And we're actually going to see Cloud Code go and write this voice agent end-to-end using the Assembly AI voice agents API and a little bit of agent instructions. Okay, so it says we're running. Let's go ahead and copy this URL and go try it in that window that we were using previously. And just to finalize this, it took about five minutes to get this done. And from a user input perspective was relatively straightforward. All right, let's go ahead and try our Taco Bell voice agent that we built with the Assembly AI voice agents API.

[00:03:06] Speaker 2: Welcome to Taco Bell. What can I get started for you today?

[00:03:09] Speaker 1: Yeah, can I get a chicken quesadilla?

[00:03:14] Speaker 2: Great choice. I've added a chicken quesadilla to your order. Would you like to add anything else? Yeah, can I get a medium Baja Blast? Perfect. Your medium Baja Blast is on the way. That brings your total to $8.48. Would you like to add anything else? Like some cinnamon twists or maybe a crunchy taco?

[00:03:37] Speaker 1: Yeah, do you have the nachos?

[00:03:44] Speaker 2: Yes, we do.

[00:03:45] Speaker 1: Wait, actually no, let's just go with the cinnamon twists.

[00:03:50] Speaker 2: No problem. No problem. Adding some cinnamon twists it is. Your order now has a chicken quesadilla, a medium Baja Blast and cinnamon twists. The total is $10.27. Would you like to add anything else or are you all set?

[00:04:05] Speaker 1: Nope, all set, thank you.

[00:04:12] Speaker 2: Perfect. Your final order is ready. One chicken quesadilla, one medium Baja Blast and cinnamon twists. Totaling $10.27. Is everything correct? If so, I'll submit your order now.

[00:04:27] Speaker 1: Yes. And so there you have it. Using cloud code and some agent instructions in five minutes we built a Taco Bell voice agent. You can see that the conversation was very natural. When I interrupted, the agent stopped talking. It let me do a run-on sentence where I said, you know, do you have the nachos? And then we also have a demo right here where it looks like the cinnamon twist was made it to the order, but not to the UI. And so there's some things here that we could of course clean up, but for building a demo or a POC, this is a really quick way for you to test the API and see how it works well for your use case. So there you have it. In five minutes, we went from agent instructions on an activation page to a fully functioning voice agent. Obviously, this is just a demo and you'll wanna do a lot more customization to bring this to production. I'd highly recommend you go check out the demo and see all the customization you could do on top of your voice agent. And if you're new to Assembly AI, go to assemblyai.com and sign up and you'll get free credits to get started. We can't wait to see all the voice agents you build with our API and are looking forward to your feedback. All right, thanks everybody. Happy building.

ai AI Insights
Arow Summary
Ryan from Assembly.ai demonstrates how to quickly build a managed voice agent using the new Voice Agents API. After signing up, he copies a pre-made prompt from the activation page into a coding agent (Cloud Code), answers a few setup questions (fresh project, managed voice agent, browser audio source, language choice, API key, US region), and the tool generates the end-to-end code automatically. Within about five minutes, he runs a browser-based Taco Bell ordering voice agent, showing natural conversation features like interruption handling and run-on speech. He notes minor UI/order-sync issues and emphasizes the demo/POC nature, recommending further customization for production and inviting users to sign up for free credits and explore customization options.
Arow Title
Build a Voice Agent in 5 Minutes with Assembly.ai Voice Agents API
Arow Keywords
Assembly.ai Remove
Voice Agents API Remove
managed voice agent Remove
Cloud Code Remove
prompt copy Remove
API key Remove
US region Remove
browser audio Remove
Twilio Remove
Telnyx Remove
Taco Bell ordering demo Remove
agent instructions Remove
proof of concept Remove
interruption handling Remove
voice UI Remove
Arow Key Takeaways
  • You can bootstrap a Voice Agents API project by copying a pre-written prompt from the activation page into a coding agent.
  • Answering a few configuration questions (source, language, region) is enough to auto-generate working end-to-end code.
  • A browser-based voice agent can be built quickly, with alternatives like phone integration via Twilio/Telnyx.
  • The demo shows natural voice interaction features such as barge-in/interruption handling and understanding run-on speech.
  • Expect to refine UI and integration details for production; the five-minute flow is ideal for demos/POCs.
  • Assembly.ai offers free credits for new signups to experiment with building voice agents.
Arow Sentiments
Positive: The tone is upbeat and promotional, highlighting ease of setup, quick time-to-demo, and natural conversational behavior, while briefly acknowledging minor issues as fixable.
Arow Enter your query
{{ secondsToHumanTime(time) }}
Back
Forward
{{ Math.round(speed * 100) / 100 }}x
{{ secondsToHumanTime(duration) }}
close
New speaker
Add speaker
close
Edit speaker
Save changes
close
Share Transcript