Introduction
At Bland, we’ve built an easy-to-integrate voice call API. Sending an AI phone call requires ten lines of code. Provide a prompt and a number to call and Bland will handle the rest. Or set up an inbound phone number - it’s even easier.
In this guide, we’ll explain why we built our API, how we built it, and how you can integrate Bland’s voice call API to send or receive your first phone call. Or, feel free to skip the blog to view our documentation or developer portal right now!
Why we built Bland AI
When most people think of automated voice calls, they think of interactive voice response (IVR). IVR systems power major banks, telecoms, and utility companies. IVR systems are slow, force callers on hold, and purposefully make it hard to reach human representatives.
What’s even worse than IVR? Robocalls. Robocalls are pre-recorded audio messages that government agencies and scammers use to plague everyday people. IVR, in concert with robocalls, has ruined the entire phone calling experience. What used to be a medium for connecting with friends and businesses, is now a channel for spam and wasting time.
In healthcare, real estate, and nearly every other industry in the world, billions of dollars are wasted annually on formulaic, repeated phone calls for data collection and sharing. Enterprises hire call centers in the Philippines, India, and South America. They burn millions of dollars, struggle to scale their operations and fail to deliver positive experiences to their customers.
We built Bland because we believe phone calls can be amazing. Phone agents can be helpful, friendly, and beyond all: competent. Our mission is to empower developers to create excellent voice experiences for their customers. Our voice call API is dead simple to integrate - let’s dive into how we built it.
How we built our Voice API
Bland’s programmable voice tool leverages three underlying AI models, plus telephone network infrastructure, to send AI voice calls.
Bland uses a transcription model to listen to incoming audio and convert it to text. Then we use a language model that intakes the text and states how the AI should respond. Finally, we have a text-to-speech model that outputs human-sounding audio. The result: a fully dynamic phone agent that’s programmable for any task.
One of Bland’s core innovations, that places its programmable voice API in a league of its own, is its phone call latency. The typical API provider or voice solution has five seconds of latency. In contrast, Bland’s voice calling API clocks in under two seconds. Send yourself a test call here.
Additionally, Bland elevates voice communication by providing a voice chat API that’s fully dynamic. Developers can inject live context into inbound phone calls. They can even call functions during phone calls, to schedule appointments, send text reminders, and update databases. As a result of its advanced features, Bland’s voice capabilities are endless. Developers can build for any use case and then quickly deploy and scale.
We might be getting ahead of ourselves though. Let’s dive into how to send your first phone call!
Sending your first phone call
Sending a Bland AI phone call couldn’t be easier. Provide your AI with a task and number to call. Then our servers dispatch the phone call, our voice agent handles the end-to-end conversation, then Bland returns a transcript from the calls. Afterward, most developers analyze the call transcript, using OpenAI’s GPT application programming interface to parse key data and formulate a response.
To start your voice call integration, sign up on the developer portal.
Then navigate to the “Send phone call” page.
Once you’re there, insert your phone number. Then, select one of the default prompt options. Or update your prompt for your phone agent to do anything from sales to customer service. Finally, update the various phone agent options and press ‘send’.
Once your phone call is sent, you’ll see a live transcript. It’ll include what the AI hears, and how it responds. After the call ends, you’ll want to quickly iterate on your prompt until it’s perfect. Lucky for you, we built the perfect tool for it.
Via Bland’s testing tool, you can communicate with your voice intelligence through a text interface. Ask your AI questions, see how it responds, and then adjust your prompt accordingly. Quickly iterate on the prompt to get the best outputs possible.
Finally, start exploring the rest of Bland’s platform. From navigating IVRs using automated machine detection software to creating voice clones to uploading CSVs of lead data and phone numbers to send large batches of phone calls, Bland can do it all.
Conclusion
Bland’s voice call API makes it incredibly easy to send and receive phone calls where an AI speaks to a human. Simply provide the AI a task, plus the phone numbers to call, and it handles the entire end-to-end conversation.
With Bland, you can automate every phone call a human being would otherwise make. You can replace your costly BPO, VAs, or any other type of outsourcing with your own in-house solution. Then you can scale up, infinitely, to millions of phone calls, the whole time only paying for usage.
You can get started by viewing Bland's documentation. Then, visit the developer portal to send a phone call or set up an inbound phone number. Finally, join our discord community to engage with and learn from our active developer base.
Until next time!