AI just called 3,000 pubs. Here’s what happened

Over St Patrick’s weekend, someone built an AI voice agent with a Northern Irish accent and set it loose on a very specific task. It called more than 3,000 pubs and asked one simple question: how much is a Guinness?

The goal was to build a price comparison site, but the method is what makes the story interesting. Instead of scraping websites or collecting information manually, the AI handled the interaction itself. It made the calls, asked the question, and gathered the answers.

There’s something slightly mischievous about it. The AI reportedly used a cloned voice, based on someone from The Traitors, adding a playful edge to the project. But underneath the novelty, the mechanism was simple: automated small talk with a clear outcome in mind.

It’s not flashy AI. There’s no complex interface or dramatic reveal. Just a simple question, asked at scale, in a way that feels surprisingly human.

But there’s a bigger shift sitting underneath it.

AI is no longer just generating content. Increasingly, it is starting to act: making calls, asking questions, collecting information from the real world, and doing so on behalf of someone else. The interaction itself becomes part of the system.

That is where things become more interesting, and slightly more complicated.

Did the pub staff know they were speaking to an AI? Would they have responded differently if they did? And at what point does something like this move from clever to intrusive?

In this case, it is fairly harmless, even inventive. But the idea scales easily. If one person can make thousands of calls over a weekend, the line between useful automation and unwanted noise starts to blur.

And that raises a slightly different question.

If AI can act on our behalf in the real world, where do we draw the line between efficiency and intrusion?