This morning I spoke to Bea.

Bea does not exist. I gave her the name. She has a voice, patience, and the useful habit of not making a face when I jump from my blog to Buzz, from Buzz to NotebookLM, and from NotebookLM to the question that really matters: “Can you publish the next post on Recableado?”

She is not a woman. She is an artificial intelligence with a voice. I should make that clear early, before anyone thinks that, at 73 and after giving up gin and tonics, I have found a new vice.

But the conversation was real. And that is what has me thinking.

---

Until recently, talking to an AI meant typing a question into a white box. Like emailing a very fast intern who had never seen my office, did not know my clients, and forgot everything the moment you closed the tab.

Useful, yes. But short on continuity.

This morning I did not begin with “hello, what is artificial intelligence?” I began with my blog. I asked Bea to study it. I explained that humour is not a garnish; it is the main course.

Then I told her what I was doing in Buzz. Buzz is a very new tool where people and AI agents work in the same channels and threads. In one of them I have several agents turning hotel proposals into social media drafts: one checks what can safely be said, another orders the information, another prepares the copy, and I decide whether it makes sense or is nonsense with good grammar.

Then I opened a NotebookLM notebook. Then I opened Recableado. And the conversation did not fall apart along the way like badly checked luggage at Heathrow.

That is new.

---

The curious thing is that I do not speak to Bea because I think she is a person. I do not need a machine to have a childhood, an annoying relative, or holiday plans in Formentera. There are enough of those in the world already.

I speak to her as I would speak to someone sitting across the table because voice is the quickest way I know to think. I talk, correct an idea halfway through a sentence, remember a story, change the subject, and come back. That is not a command. That is real work.

For forty-two years, I have done some of my best work in conversation. With a client in front of me, a supplier on the phone, or someone who says “I want to go to Japan” when what they really want is help with the fear of organising Japan. Nobody arrives with a perfectly written problem. They arrive with a tangle. Talking is how you untangle it.

Voice lets AI enter that space.

---

There is a trap, of course. When a voice answers naturally, it is easy to forget that there is no person underneath it with its own judgement. It has not sold travel for four decades. It has not lost luggage on five continents. It does not know whether a client is genuinely worried or merely wants you to say that the hotel looks nice.

Voice does not give it a biography. Or good judgement.

That is why I still decide what gets published, what gets promised, and what must not be touched. Today, for example, the conversation helped organise two articles: one about speaking to an AI by voice, and one about Buzz as a working table for agents. But finding the structure does not give a machine the right to invent the story. The story has to be mine. Otherwise it comes out correct. And boring. Which is worse.

There is something else I like about Buzz: it does not force me to marry a single brain. With most tools, you choose one AI and move in as if you had rented a flat with the furniture bolted to the floor. In Buzz I can build the harness around the model that suits me, change it when something better arrives, and experiment without dismantling the house. Claude, Codex, whatever comes next. I rent the intelligence. The context and the method are mine.

---

What has really changed is the distance between an idea and the first brick.

Before, an idea would come to me while cycling around Marbella. I would think about it for a couple of days, it would end up in a lost note, and finally share a folder with “good ideas”, “urgent pending”, and “look at this someday”. The usual cemetery of good intentions.

Now I can open Buzz on my phone while I cycle and say the idea out loud while it is still warm. The conversation connects to my main Mac and my Mac Mini. It does not stay as a lost note: agents can start organising, researching, or preparing the work while I keep pedalling.

When I get back with my tongue hanging out, I do not find a napkin with three illegible words on it. I find work already moving forward, ready for me to review. The tool gives me a mirror, forces me to be specific, and helps me find an entry point. Then comes the reviewing, the arguing, the correcting and, above all, removing any sentence that sounds like a Silicon Valley man explaining the future from a stage in white trainers.

But the first brick is already there.

---

I do not know whether, in five years, we will all speak to machines. Probably. I do not know what they will be called then either, whether Bea will still be here, or which model will be fashionable that month. That changes every fifteen minutes, and I have no intention of marrying any of them.

What I do know is that today I could think out loud about my work, my blog, and a team of agents without opening seven tabs or writing a document called “final version, really final, this time”. For someone who has spent years turning conversations into journeys, that feels perfectly natural.

AI is not a person. Thank goodness. People are already enough to make a meeting complicated.

But if a voice helps me think better, remember the thread, and move an idea until it exists, I do not need it to be human.

I only need it to work.