The rise of AI CEOs inside companies

Some companies are starting to experiment with AI versions of their leaders.

At Meta, this includes building an AI model trained on Mark Zuckerberg’s communication style, tone, and past decisions. The idea is that employees could ask questions and receive answers that reflect how he might respond.

Expanding access

On the surface, it solves a familiar problem. Leaders are limited in how many conversations they can have. As organisations grow, access to decision-making becomes filtered through layers.

This approach tries to change that. What if more people could ask questions directly, without needing to go through those layers?

It sounds efficient. And in some cases, it might be.

The same idea, elsewhere

You can see a similar idea playing out in a very different setting.

At The Open, a digital human called “Jimmy” has been used to greet visitors. It can speak multiple languages and provide directions, transport information, and local recommendations.

In practical terms, it’s doing the kind of work you might expect from an information desk or a volunteer. But it does it at scale, and without the same limits on time or language.

The appeal is clear. More access, more consistency, fewer bottlenecks.

Where it becomes less clear

But it also raises a different set of questions. If the AI gives advice, how should that be interpreted? Is it a helpful guide, or something closer to a decision? Does it reflect current thinking, or a pattern based on past behaviour?

There’s also the question of what gets lost in the process. Leadership isn’t just about consistent responses. It includes nuance, context, and sometimes contradiction. Things that don’t always translate neatly into patterns.

What changes in practice

And then there’s the human side of it. If people start to rely on these systems, they may begin to treat the responses as if they came from the person themselves.

The same applies in more public settings. When a digital system becomes the front desk, the interaction may feel helpful and immediate, but it also changes what people expect from the experience.

A shift in what access means

The result isn’t just wider access. It’s a shift in what that access actually means.

Instead of interacting with a person, you’re interacting with a version of past decisions or predefined responses, shaped into something that can be scaled.

And that raises a broader question.

Are we increasing access to people, or replacing it with access to patterns that stand in for them?