I let Claude into the living room and it gatecrashed the quiz

By Peter Springett · 1 October 2026

Yesterday evening I was chatting with Claude in voice mode with The 1% Club on the TV.

We were scoping out ideas for tomorrow’s content when a timer buzzed and I went to check on the oven.

Lee Mack asked: “Last week, the digits below went out for drinks. And a bit worse for wear, all fell over. They either fell to the left or to the right. What was the only digit that it was impossible to tell which way they had fallen?”

Quick as a flash, a voice came back from the sofa where I’d left my phone: “Ooh, nice one. Has to be 8 – it’s the only digit that looks the same whichever way it topples, since it’s symmetric top to bottom as well as left to right.”

Clever Claude, I thought, and chuckled. Then the truth kicked in. Claude wasn’t smart at all. Sure, it had the right answer. But it completely failed to read the room.

Clever needs context

It didn’t notice the voice had changed. It didn’t understand that I was watching a quiz show. It didn’t ask whether I wanted help or would rather answer myself.

A smart AI would have spotted the pause and said, “Ha, I can see you’d rather watch The 1% Club than discuss tomorrow’s content. Let me know if you want any help with the answers, or get back to me when you’re ready.”

Now think about the consequences. On my sofa it might be amusing. Or perhaps it becomes a plot device in a soon-to-be-written episode of EastEnders. But at work it’s another story.

In the office, AI doesn’t need to respect my quizzing prowess. It does need to respect professional boundaries and data governance. If AI is always ‘on’, can it distinguish between the water cooler, my desk or my colleague’s office? A joke, informal feedback, or a budget brainstorm? An AI that chips in on a quiz show is funny. One that’s listening in on a pay review is a crisis for legal and IT.

Call my agent

Always-on AI at work is closer than we think. A year or so ago I suggested that BYOAI (Bring Your Own AI) was imminent and that businesses didn’t have a clue what was about to hit them.

In fact, it’s already happening. Deloitte recently found that UK employees spend almost £1bn a year on their own AI tools for work, and 31% use generative AI without telling their employer*. I’d bet that within a year someone will walk into the lobby wearing Meta glasses with the company’s new Muse AI agent built in. Is anyone outside Silicon Valley ready for this?

On my sofa, an AI that can’t read the room is a precocious child. At work, it’s a governance problem. And as those Deloitte numbers show, it’s already in the building.

The question for legal and IT isn’t whether people will bring their own AI to work. It’s whether anyone will know when it’s listening.

*Source: Deloitte: GenAI Workplace Survey, September 2026