Death of the TUI
Press '1' to enter the modern world
What’s in a TUI?
You might not recognize the acronym TUI. But you’ll have used a TUI (or Telephone User Interface) many times over the years. They are the frustrating things which ask you to ‘press 1 for sales, 2 for support or 3 to get lost in a myriad of extra options’. Sometimes you get lucky and repeatedly pressing ‘#’ will let you talk to a human. But more often than not minutes of non-joy lie ahead listening to depressingly cheery hold music.
The future
But AI is set to change this. Conversational UIs - CUIs - will increasingly replace TUIs. Soon, when you call your doctor you’ll talk to an AI receptionist. They’ll recognize you (using a combination of caller ID and voice recognition), arrange appointments, remind you of prescriptions you need to collect and more. They’ll offer a personalized, friendly service. They’ll scale out to meet demand - no more listening to music while you are stuck in a queue.
And it gets better. Just before Christmas, OpenAI integrated ChatGPT with WhatsApp - add ChatGPT to your contacts and you can WhatsApp with ChatGPT.
Rather than having to dial a phone number, these AI receptionists will be available through the messaging apps you already use. You’ll be able to schedule appointments through WhatsApp or Signal. Call or message. Take your pick.
This has another benefit - security. Traditional phone systems use protocols dating from the 1970s that lack basic authentication features. Known as SS7, these protocols were designed when telecoms were closed national monopolies, with security through obscurity. But that world is long gone. In recent years, security researchers have shown how SS7's lack of authentication can be exploited to track locations, intercept calls and messages, and even hijack phone numbers. These vulnerabilities have already been exploited by foreign actors to spy on individuals. Modern messaging apps provide better security through end-to-end encryption and strong authentication. We get a better user experience and better security.
New possibilities
The integration with messaging platforms opens up new possibilities. Your AI receptionist could proactively remind you about upcoming appointments or when prescriptions need renewal. Rather than waiting on hold to speak to a doctor, you could have an asynchronous chat conversation when it suits you. The AI could even help prepare you for appointments by gathering preliminary information or symptoms.
This integration with modern messaging platforms solves another key problem - identity verification. When you call your doctor's office today, they typically verify your identity by asking for your date of birth or address. But with a messaging platform like WhatsApp, your identity is already verified through your phone number and the app's security features. Trust is established from the start.
The shift from phone calls to messaging also means the interaction can be richer. The AI can share links to medical forms that need completing, send maps with directions to the clinic, or even share educational content about managing your condition. You could send photos of symptoms or share previous medical records directly through a secure chat. If you’ve got a health app on your phone AI can send summarized information direct to your doctor.
For healthcare providers, this approach offers significant advantages. They can handle more patient interactions simultaneously since asynchronous messaging doesn't require dedicated staff time like phone calls do. The AI can handle routine tasks 24/7, freeing up human staff for more complex cases. And there's a clear audit trail of all interactions, handy in a highly regulated industry.
Not just for doctors
And it's not just healthcare. Imagine the same convenience when dealing with your lawyer, estate agent or bank. Instead of playing phone tag or scheduling meetings for simple updates, you’ll message their AI assistant through WhatsApp. Need an update on your house purchase? The AI will instantly provide the latest status and share important documents. Looking for a new home? The AI will learn your preferences and proactively notify you when relevant properties come on the market. Any service provider that currently relies on phone-based customer service will benefit from this shift.
Managing the risks
However, this transformation isn't without challenges. There’s the risk of jailbreaks - getting the AI to inadvertently reveal confidential information (although remember humans aren’t infallible either). And AI artefacts. Poor early implementations will damage trust (e.g. Apple AI message summaries), so handing over to humans early on will be essential, especially in critical situations. There may even be a role for oversight agents which monitor human-AI calls to spot when they go off-piste.
The future
The days of navigating complex phone menus or waiting on hold are numbered. Instead, we'll interact through familiar messaging platforms with AI assistants that know who we are, understand our history with their organization, and can provide personalized, efficient service.
Remember those minutes of non-joy listening to depressingly cheery hold music? They'll become a relic of the past, like rotary phones and phone books. Because ultimately, we humans just want our problems solved quickly and reliably. If AI can do that while providing better security and convenience, we'll wonder how we ever put up with ‘press 1 for sales’ in the first place. The end of the TUI era is approaching - and it can't come soon enough.


I think you mean death of the phone - not just of the TUI. Just another nail in the coffin for the telephone. Why would I want to talk to either an AI or a person on the phone if I could do what I needed to do any other way? Sure I still want to talk to my friends or work contacts but I’ll use Zoom, Whatsapp, whatever for this, not the phone. And for the things I need to use the phone for - many of which involve TUIs - I have no desire to ever pick up the phone for them.
Also, how many TUIs, or how much of a TUI’s implementation, is there to stop us getting through to someone in the first place? They’re not particularly trying to be helpful, but rather prevent themselves from having to spend money dealing with you. So why bother implementing a helpful AI agent if it’ll cost money to implement and run?
Oh the cynicism…