A new network-based agentic artificial intelligence helper is being tested with customers
At AT&T, we want to help customers spend less time distracted by unwanted phone calls, so they can spend more time connecting with the people they want to talk to.
AT&T has a long history of protecting customers from spam and fraud. We currently block or label more than 2 billion robocalls per month. Spammers are getting more sophisticated, and they are using AI to spam people even more. So, we’re fighting fire with fire by trying out the latest weapon in the battle against fraud and spam: agentic artificial intelligence.
We’re testing a new AT&T digital receptionist, which is rolling out to select customers throughout this year.
The digital receptionist adds on to our other spam filters such as AT&T ActiveArmor® by literally talking to callers to determine which calls are suspicious. We’ve taken years of data and knowledge about how spam and fraud works – and our algorithms for fighting them – to create this new powerful tool.
It uses advanced voice-to-voice and agentic AI to screen your calls, helping you only take the calls you want to take. That means you can spend more of your time on the calls that are important to you, and let AI handle the rest.
This is just a taste of what AI agents could do with your phone to make your life easier. Eventually, a future AI agent could autonomously connect you to make reservations at the hottest restaurant. Just tell it the name of the restaurant, and it will connect you – or even book your reservation – all on its own. This digital receptionist could also eventually handle phone calls that are important, but you can’t take at the time, by getting directions from you in real time with text prompts.
How will the digital receptionist work?
We’re still testing features and adding capabilities. But here’s a look at what’s possible.
Just like a human receptionist, the digital receptionist sits between you and the outside world. It answers an incoming call on your behalf, prompting questions like: “Who may I say is calling?” and “What is this in regard to?” It determines whether the caller is human, how urgent the call is, and if it meets your customized criteria. If the call passes all the checks, the digital receptionist lets the call through to you and drops off the call completely. If the call is about something that the AI can handle on its own, like taking a message or accepting a delivery window, it will.
If the caller won’t identify themselves, if they called the wrong number, or if the call doesn’t meet your criteria, the receptionist will either disconnect the call or take a message. While the digital receptionist is talking to the caller, you can watch a live transcript and pick up at any time. Or, let the receptionist handle the call, then read a text summary and decide whether to call back yourself.
Calls from trusted numbers, like your friends, family, or doctor, can be put on a “Do Not Screen” list, and are automatically put through with no AI involvement.
The digital receptionist could talk to incoming callers to screen for spam, allowing customers to spend more time connecting to the people they want to talk to.
HTML Editor Component
*Contents may not have visible height
Reduced Width Video Banner Code
What makes this unique?
You might say, “I’ve seen this from other companies.” But ours is built differently to make it easier and more convenient for our AT&T customers.
We’re uniquely positioned to build this directly into the network itself to work with your phone number. It’ll work without any cumbersome downloads. It won’t drain a phone’s battery. And it’ll still work even if you’re out of cellular service range or not connected to the network.
This will be one of the telco industry’s first agentic voice AI tools used directly by the customer, leaving other carriers in the dust when it comes to integrating consumer-focused agentic AI tools into the network. This opt-in network-based agentic solution could make decisions based on your parameters and input, not just search for information and create content.
What’s the tech behind this?
We use multiple large language models (LLMs) to process the incoming speech, create responses, and turn those responses back into speech. All this work is done with the same responsiveness you’d expect from a human, making the conversation with the digital receptionist feel natural.
We used decades of expertise in fraud prevention building spam-fighting algorithms to inform how this digital receptionist detects suspicious calls. If it recognizes the patterns of spam or fraud, it shuts down the call automatically. It’s all aimed at one goal: connecting you to what matters most.
And it’s all governed by our strict security and privacy policies. Your information is safe and secure, and only used to benefit you by helping the receptionist tailor its experiences to your preferences.
What does this mean for the future?
Putting our customers first is at the core of everything we do, and we’re constantly exploring new ways to make that happen.
We’re working on lots of agentic AI use cases, which we’ll be talking about as soon as we can. While this digital receptionist is just in its testing phase for now, we’re drawing on our strong track record of innovation and AI expertise to create experiences you would expect – and some you might not – from cutting-edge technology like AI. We’ll continue sharing updates as we roll out new AI tools to increase efficiency, convenience, and put our customers first, period.
