The most advanced AI in the world won't connect if it can't communicate like a human. Before ChatGPT became a household name, Google brought me on board to help shape how their AI would speak to people.

The real challenge isn't teaching machines to talk - it's helping them find a voice that feels authentic. Every prompt, every response, every conversation needs to strike a delicate balance between being helpful and being human.

Working directly with Google's AI team, we developed conversations that felt natural without pretending to be human. This meant understanding not just what AI should say, but how it should say it - from explaining complex concepts simply to knowing when humour helps (and when it doesn't).

Now I'm helping other organisations figure out their AI voice, creating bots that actually engage people.

I’ve even built a tool to help others define their bot's tone of voice. It's very meta - a bit like teaching AI how to teach AI how to talk. (Sorry, couldn't resist that one).

The work helped shape how Google's AI communicates with billions of users. More importantly, it showed that artificial intelligence can have authentic conversations - when you focus on finding the right voice.

The work continues as more organisations grapple with how to make their AI sound more human without losing trust. It's about finding that sweet spot between capability and connection. After all, even the smartest AI needs to know how to read the room.