We called it science fiction. Well, that science fiction has arrived, quietly. And most leaders I know are still typing.
We met AI through a keyboard
When ChatGPT launched in November 2022, most of the world’s first real conversation with AI happened in a text box.
A whole discipline grew up around it: prompt engineering. We learned to write short, precise, carefully built instructions. Role, task, format, constraints. We traded templates. We treated the prompt like a telegram, where every word cost something.
That made sense at the time, but it came at a price. When you type, you edit before you’ve finished thinking. You leave out the context that feels too obvious to write down, the caveat that takes too long to explain, and the “here’s what actually worries me” that never makes it into the box.
The AI only knows what you give it. In the text era, we gave it a summary of our thinking, not the thinking itself.
The interface is shifting to voice
That is changing fast. Voice is now built into the AI assistants and agents many of us use every day. I mean conversation not dictation: real-time and back-and-forth. You can interrupt, change your mind mid-sentence, and the AI follows along.
Why voice changes the quality of the work
Bandwidth. Most people speak at around 130 to 150 words per minute and type at around 40. In the same three minutes, you can give your AI three times as much to work with, with less effort.
Range. Speech carries things that text strips out: emphasis, hesitation, the moment you say “actually, wait, here’s what I really mean.” Those signals tell the AI what matters most and where you’re still unsure.
Rambling is a feature, not a bug. This is the part I find most important. In the text era, brevity looked like a virtue. It was really a limitation of the keyboard. When you talk, you ramble, and in that ramble you naturally supply the background, the constraints, the politics and the priorities you would never bother to type.
Today’s AI is remarkably good at taking that messy, honest stream of thought and finding the structure in it. You don’t have to organize your thinking before you hand it over. The AI can help you organize it.
The best prompt isn’t the shortest one. It’s the most complete one.
Leaders already lead by voice
Here is why I think this matters especially for leaders.
Think about how you actually run your organization. You don’t send your leadership team a perfectly formatted prompt. You brief them. You talk through the goal, the context, what good looks like and what to watch out for. You think out loud in one-on-ones. You redirect in the hallway. Leadership has always been a spoken discipline.
Now look at where AI is heading: agents. These systems don’t just answer questions. They take on tasks, work across tools and come back with results. Working with an agent feels less like using software and more like delegating.
And leaders delegate by talking.
Voice also changes where and when the work happens: walking between meetings, driving, sitting in the car before you walk into the office. Michael Knight had it right. Some of your best direction-setting happens on the move.
The honest caveats
None of this means voice wins every time. A few things are worth thinking through:
- Not every task suits voice. Precise edits, numbers and code are often still better typed. The goal is to pick the right mode for the task, not to abandon typing all together.
- Setting matters.Open-plan offices and confidential topics don’t bode well with talking out loud to your laptop.
- A voice is not an identity. Voice cloning is real and improving fast. A spoken instruction alone should never authorize a payment, a contract or access to sensitive systems. If your organization is deploying voice-enabled agents, authentication and governance come first.
- Thinking out loud takes practice. Many professionals spent decades learning to speak only once the thought was finished. Rambling to an AI can feel awkward at first. That feeling fades quickly.
The lesson: change the habit, not just the tool
So here is my challenge for this week. Pick one task you would normally type to your AI, and talk instead. Give yourself three minutes. Ramble. Include the context you’d normally leave out. Then compare the result with what you usually get.
If you lead a team, do it where they can see you. Your habits set theirs.