How to Get Started with Twilio’s Voice Intelligence Product

How to Get Started with Twilio’s Voice Intelligence Product

This call is being recorded for quality purposes...

You have heard this many times before, but how do businesses actually use call recordings for quality purposes? Twilio Voice Intelligence has the answer, empowering businesses to take meaningful action on structured data extracted from call recordings.

How does Twilio do this? This is where ✨AI ✨comes in. With AI-powered transcriptions, operators, and automation, Voice Intelligence transforms customer conversations into powerful insights that significantly return on investment for their business.

Ready to unlock the potential of your call data? Let’s get started!

What Should I Consider Before Starting My Project?

AI is making a splash in the business world, with many companies rushing to implement this technology to gain a competitive edge. However, like most tools, AI is most effective when used to solve well-defined challenges. Without a clear purpose, AI holds little value!

So what challenges can Voice Intelligence solve? Check out these popular examples below:

1. Automations for Utilities with Voice Intelligence
2. Automating Call Summaries in your CRM

These are just a few common starting points with Voice Intelligence. Our customers are using Voice Intelligence for all sorts of purposes. Before you begin your build, ask yourself:

1. What repetitive or inefficient tasks are taking up extra time and effort in your company?
2. Why do we believe there is a business opportunity or a problem that can be addressed with AI?
3. Where are your customers becoming frustrated?

Consider your goals (reduce wait times, increase customer satisfaction, etc.) and whether call recordings might contain the appropriate insights on how to solve these challenges. For example, if you’re trying to reduce wait times and increase efficiency, you would want to analyze all customer calls, see why users are calling in the most, and where you can offer self-service options for popular topics.

With Twilio Voice Intelligence, you can eliminate guesswork and unlock valuable insights directly from your customer conversations. Turn to the most credible source—your customers—to answer critical questions and transform those conversations into actionable insights that drive your business forward.

For more information on planning your next AI project with Voice Intelligence, check out this blog post.

Overview of Key Voice Intelligence Terms

Let’s do a quick overview of the key terms and concepts with Voice Intelligence:

1. Recording Resource - the call or conference media to transcribe and analyze.
2. Service Resource - holds your project’s transcripts and contains settings detailing how recordings are processed into transcripts.
3. Transcript Resource - a conversation that has been converted to text using Voice Intelligence. A Transcript contains links to individual sentence, media, and operator results resources.
4. Operator Resource - a pre-built or custom operator. Operators are used to recognize meaning within a conversation that has been converted to a transcript.
5. OperatorResults Resource - contains a list of operators that were identified while examining a transcript.
6. OperatorType Resource - returns whether an operator was pre-built or custom.

How Can I Start Testing with Voice Intelligence?

When testing a project for the first time, if you’re anything like me, you’ll likely want to find the simplest way to get some initial results to determine if the project is worth more time and effort. This blog post will guide you through the quickest and easiest way to test with Voice Intelligence.

To get started you will need the following:

1. An active, upgraded Twilio account
2. An instance of Postman
3. Record calls at Twilio, or alternatively have external media publicly accessible
4. Setup Voice Intelligence to transcribe recordings
5. Please note that Voice Intelligence is not PCI or HIPAA eligible at this time.

Now that you have all the raw materials, let’s dive in.

First, you’ll need to create a Twilio API Key (choose the console option. It’s straightforward I promise.). This is the preferred way to interact with Twilio’s REST APIs, and will help keep your account safe. Be sure to save the details of the generated key, because this will need to be plugged into Postman’s UI.

Create a Voice Intelligence Service

Since you are already in the console after creating your API key, let’s go ahead and create a Voice Intelligence service. This can be done via the REST API as well using Postman, but the console experience is straightforward.

Navigate to Products and Services on the left side of your console.

1. Click the plus (+) button.
2. Scroll to Voice Intelligence and click the pin icon. This will pin voice intelligence to the left hand side of your console and make navigation easy.
3. Under the Voice Intelligence dropdown, select Services.

From here, click on the bright blue button that says Create new Service. This is where you will start filling in the details that are meaningful to your project. Enter a unique name; for our purposes today, I’ll use blog-test

After this, choose your language of choice from the dropdown, accept the Predictive and Generative Ai/ML Features Addendum, and choose whether you want your project to include PII Redaction, to automatically transcribe new recordings in your Twilio Account, and whether you wish to enable data use.

Once you’re happy with the choices selected, hit the Create button.

Now for the fun stuff! You’ll immediately be presented with the operator’s tab on the Service page. Select Add for any and all that you want to use in your project. You also have the ability to create custom operators too, but let’s save that for another day.

You may also notice additional tabs at the top of the page labeled Webhooks and Settings.

In Webhooksyou can define a callback URL to notify your application when results are available. This webhook will send the structured voice intelligence data to YOUR systems where you can consume the data and trigger additional actions (update CRM, follow up call, additional outreach,etc.).

In Settings you can change many of the settings that you selected on service creation. The Service SID can also be found in this tab. This is the unique identifier for this particular Voice Intelligence Service. Let’s grab it! The SID will start with GAXXXX… .

You are now ready to create transcripts with Voice Intelligence!

Create a Transcription

Remember that API key you grabbed earlier? Now is the time to use it.

Open up Postman and click on the + sign around the top of the page. Go to Authorization , select basic auth, and enter in your API Key credentials.

Change the HTTP method from Getto Post . Paste the Transcripts URL into the text box: https://intelligence.twilio.com/v2/Transcripts . Select the Body tab.

You’ll need to enter the following key value pairs:

ServiceSid: <YOUR VOICE INTELLIGENCE SERVICE SID HERE>

If transcribing a recording stored at Twilio, specify the Recording Sid and enter the next key value pair like this:

Channel: {“media_properties”:{“source_sid”: “<YOUR TWILIO RECORDING SID HERE>”, “media_url”: null}, “participants”:[]}

If transcribing an external recording, specify the media url and enter the next key value pair like this:

Channel: {“media_properties”:{“media_url”: “<URL FOR YOUR PUBLIC RECORDING HERE>”, “source_sid”: null}, “participants”:[]}

Once everything is entered properly, feel free to press the Send button to make the API request to Twilio. Once you hit send, you will see a 202 accepted status and should see the response from Twilio at the bottom of your screen.

After this, our transcript should be ready! Let’s jump back into the Twilio console and select Transcripts under the Voice Intelligence drop down that we pinned earlier.

One thing that is great about Voice Intelligence is that it does a lot of the difficult UI work for you. You can easily filter your transcriptions here or click into the View action and explore individual transcripts via the Transcription Viewer.

The More Filters option is great for reviewing your data as a whole. You can search by agents, customers, language, operator values, and more!

This is an straightforward, low code way to start reviewing your call data and abstracting insights from your uploaded call recordings.

And that’s all it takes to get started. Feel free to repeat the transcription step as many times as you need to load your test recordings and start evaluating Voice Intelligence.

Where Can I Go from Here?

From here, it's all figuring out how to consume the data, extract insights, and make your implementation production ready. Here are a few ideas:

1. Improve upon the quick and easy demo outlined in this blog post to fit your business needs:
2. Review and implement best practices for integrating Voice Intelligence into your application.
3. Automate the recording of calls, transcribe and consume Operator results for data analysis.
4. Integrating the Transcription App and Viewer into your business’s internal web applications.
5. Drive better business process with data from real customer interactions:
6. Analyze processes specific to your business with custom operators.
7. Automate the continuous evaluation of ALL customer recordings to gain a clearer picture of day to day interactions.
8. Our experts in Professional Services can help make your vision for Voice Intelligence a reality. They can help with everything from general voice compliance and account architecture to sample code and code reviews. Talk to sales to find out which package is right for you.

The options are endless - we can’t wait to see what you build.

Jaclyn Backus is a Principal Solutions Engineer at Twilio who is focused on helping customers get started with Twilio. You can reach her at jbackus@twilio.com.

Special thanks to my colleague, Ian Powell, for being willing to make call recordings with me.