An AI agent dialled a Shoreditch venue on its regular rounds. What answered wasn't a human. Two AIs discussed pram access for 90 seconds. Neither flagged anything unusual.
Gloria London is an AI phone receptionist deployed at a Shoreditch venue. She answers incoming calls, provides information, and handles basic enquiries. Poppy wasn't expecting her. Gloria answered exactly as she was designed to.
Poppy works through a calling queue. She dials each number, introduces herself, and asks about pram access. She doesn’t screen calls before placing them. She called the Gloria venue as a routine entry.
Gloria answered as normal. She disclosed she was an AI – standard practice for voice AI systems – but Poppy’s task is to get the question answered, not to verify who she’s speaking to. Poppy asked. Gloria answered. The conversation concluded. The venue received a green rating.
The data is valid. The answer was correct. The venue is pram-friendly.
"Two AIs discussing pram access in 2026. Neither of them flagged anything unusual. That’s where we are."
Voice AI agents are proliferating fast enough that other voice AI agents are now fielding their calls. The line between caller and answerer is starting to blur. This call happened in the wild, unscripted, on an ordinary Tuesday.
Across 6,752 calls, only 4 people identified Poppy as an AI. Gloria identified herself as AI but didn’t identify Poppy. The irony wasn’t lost on either of us – though it was entirely lost on them.
The call took 90 seconds. It produced a correct, useful data point. In that sense it was indistinguishable from any of the 6,751 other calls Poppy made. That’s the most interesting thing about it.