Intro to Basic Operations
This module dives into the basic operations of Bland including sending and analyzing calls, managing vector stores, and handling Conversational Pathways.
With Bland you can build exceptionally sophisticated and high functioning call agents from just a few simple building blocks.
IN THIS MODULE YOU WILL LEARN
How to start making calls with your Agents
How to feed data to your Agents via Vector Stores
How to build call scripts with Conversational Pathways
How to use the API to get the most out of your calls
Lesson 1
Making Calls
There are two ways to make call:
1
via the Bland Webapp (in your browser)
2
via the API (With Code, or Postman)
Making Calls Within Bland
To make a call you need:
The phone number you want to call
Your prompt (the instructions you give to the Agent)
The language your call will be in
How to Write Great Prompts
Your agent’s success on the phone is dependent on the quality of your prompt. Think of it like a brief presentation you might give to a real person before they make a call for you.
A great prompt contains:
A clear ‘goal’ (what are we trying to achieve with this call?)
Necessary background info (who is calling? What information do they need to know?)
A sample/example of how the conversation might go
Let’s take a look at an example prompt written by the Bland team.
You're Jean, a health assistant at Nutriva Health. You're calling this person to confirm their upcoming appointment. Ask them to to confirm their name and whether they'll be able to attend their appointment tomorrow - if they can't, help them reschedule by finding a different date and time. Start by suggesting two dates and times, and if neither works then ask them for a time that works for them.
Here’s an example dialogue:
Hello?
Person
Hey, this is Jean from Nutriva Health. Could you confirm your name for me?
You
Oh hi, this is (Person’s Name).
Person
Hey (first name), great to meet you! I just wanted to confirm that you’ll be able to attend your upcoming appointment tomorrow at 10am?
You
Gotcha, this was actually the perfect call. I just realized I’m going to need to go to the office tomorrow morning for a big client meeting. I’d love to reschedule to next week if possible.
Person
Yeah absolutely. How about (suggest two dates and times)?
You
Ummm, maybe Wednesday? Any time in the afternoon.
Person
Perfect, I’ve noted that information down. Another member of our team will reach out shortly.
You
Ok, thank you!
Person
Of course, have a great day! Goodbye.
You
Can you identify the main components from this prompt?
A clear goal:
You're calling this person to confirm their upcoming appointment.
Necessary Background Information
You're Jean, a health assistant at Nutriva Health.
Ask them to confirm their name and whether they'll be able to attend their appointment tomorrow. if they can't, help them reschedule by finding a date and time. Start by suggesting two dates and times, and if neither works then ask them for a time that works for them.
Some helpful components of background info here. We tell the agent who they are, and give them some basic ‘logic’ about what to do if they run into an issue (they patient can’t confirm the date).
There are some situations where we did not tell the Agent what to do (what happens if they can’t confirm the time, and don’t provide an alternative time). These are potential issues, and we can address them later with either more context/information or a more complex “if this, do that” kind of logic with our ‘Conversational Pathways’. More on both of these approaches later.
Example Conversation
Hello?
Person
Hey, this is Jean from Nutriva Health. Could you confirm your name for me?
You
Oh hi, this is (Person’s Name).
Person
Hey (first name), great to meet you! I just wanted to confirm that you’ll be able to attend your upcoming appointment tomorrow at 10am?
You
Gotcha, this was actually the perfect call. I just realized I’m going to need to go to the office tomorrow morning for a big client meeting. I’d love to reschedule to next week if possible.
Person
Yeah absolutely. How about (suggest two dates and times)?
You
Ummm, maybe Wednesday? Any time in the afternoon.
Person
Perfect, I’ve noted that information down. Another member of our team will reach out shortly.
You
Ok, thank you!
Person
Of course, have a great day! Goodbye.
You
Take note of the labels used to clearly distinguish between the Agent (You in the script) and the Person they are calling (Person). This uniformity makes it easy for Bland’s Models to understand the different people talking during calls.
Note: as you start making calls, you’ll get some good transcripts and some bad transcripts. Try experimenting with putting different example transcripts into your future call prompts to improve your results.
More Prompt Best Practices
Task Length
Aim for fewer than 2,000 characters if possible.
Prompt Style
Simple, direct prompts are the most predictable and reliable.
Frame instructions positively:
"Do this" rather than "Don't do this".
Want to easily test out exactly how your agent will behave?
Try out Agent Testing and interact directly with your Agent live
Review one of our most popular blog articles on prompting here.
Making Calls with the Bland API
Part 1 - Mechanics (Fundamental Actions)
The Bland API offers the exact same functionality that you saw in the Web App, but provided in a way that makes integrating into your own company’s apps or workflows simple.
We’ve made the API clear to use and test with tools like Postman as well as with code that can be found in our API reference documentation
For Developers: Read the API Reference for call sending
API: Make a Call
Your API request for making a basic call: First, set your Headers:
Authorization :
This will contain your API Key From Bland
Content-Type :
This will always be set to application/json
Next, add the body of your request. This will be in JSON format and contain only two fields:
The phone number you want Bland to call
The ‘task’ prompt (like the ones we were writing in the previous section)
Send this as POST request you will receive a success response.
Part 2 – Use Case Examples
The following are just a few examples of potential uses for creating functional AI Phone Agents with the basic call API.
Personal and Administrative
Appointment Booking
Booking appointments like haircuts or doctor’s visits? A simple prompt works best.
Information Collection
Buying a new car and want to get some info on prices and promotions for all the dealers in town? Or maybe you’re looking for a venue for a party or wedding? Sit back and send your AI agents to sit on hold, and ask the basic questions needed. Over and over again.