We've been trying to move past email.

We replaced it with chat, collaboration tools, and faster, more structured ways to communicate. Tools like Slack and Teams didn’t just improve communication, they changed the expectations around it. Responses became quicker. Conversations became more fluid. Talking started to feel more like texting than sending a formal message.

Email, in comparison, began to feel heavy.

It required more thought. More structure. More intention. You didn’t just send an email, you composed it. You considered tone, clarity, and completeness. Even small exchanges carried a kind of cognitive weight.

So we moved away from it.

But in doing so, we weren’t just replacing a tool. We were abstracting around a limitation.

The friction in email wasn’t just the interface.

It was the people on the other side.

Progress through email was slow because humans are slow. Responses lag. Context gets lost. Artifacts take time to produce, review, and return. Even when everything is clear, there’s still a delay between asking for something and it actually moving forward.

That delay shaped how email was perceived.

It made communication feel like work. It turned even simple exchanges into tasks. Over time, that cognitive overhead became the thing people were trying to escape.

Slack channels didn’t solve the underlying structure of work. It made the interaction feel lighter. Faster replies, shorter messages, less formality. The friction wasn’t removed, it was softened. It made talking easier. It didn’t make execution clearer.

Messages became streams. Requests became conversations. Outcomes became less defined.

That works when the goal is alignment between people.

It’s less effective when the goal is getting something done.

What changes with AI isn’t just speed.

It’s where the bottleneck lives.

When the actor on the other side of a request is no longer a person, the two biggest sources of friction in email start to disappear.

The first is execution latency. An agent doesn’t delay. It doesn’t need hours, days, or weeks to respond. The gap between request and action collapses.

The second is cognitive overhead. You don’t need to carefully compose something for an agent. You don’t need to manage tone, phrasing, or social nuance. The interaction shifts from communication to intent.

You stop writing emails.

You start issuing requests.

If AI makes communication faster, why not lean fully into chat? Why not use instant messaging or a conversational interface where feedback is immediate and the interaction feels fluid?

On the surface, that seems like the right abstraction. If the bottleneck is gone, shouldn’t everything become real-time?

The problem is that chat and email carry fundamentally different mental models.

Chat is optimized for queries. Short, quick asks that can be answered in isolation. You ask, it responds, you move on. It assumes the task can be completed in a single exchange or a tight loop of back-and-forth.

Email is structured for delegation.

It assumes something larger is being handed off. Something that may require assessment, tool use, orchestration, iteration, and eventual return. You’re not trying to one-shot an outcome. You’re opening a thread for work to move forward over time.

Even with AI, that work isn’t always instant.

As systems become more capable, the tasks we delegate to them become more complex. More context, more steps, more orchestration. The interaction speeds up, but the shape of the work doesn’t collapse into a single response.

On the surface, nothing about email has changed.

But the protocol underneath it carries properties that become more valuable in this new context.

Email is identity-native. Every message has a sender and a recipient, without needing additional layers of authentication or coordination.

It’s asynchronous by default. There’s no expectation of presence, no pressure to respond immediately, but no restriction on speed when a response is available.

It handles multi-modal input without structure. Attachments, links, text, all passed together in a single unit.

And it’s universal. It doesn’t belong to a platform. It exists across all systems as a shared layer.

These were useful traits before.

With AI, they become foundational.

Email didn’t fall behind.

We moved away from it because we optimized for how humans prefer to communicate. Faster, lighter, and less formal.

But that preference was shaped by the limitations of human-to-human work. As those limitations change, so does the value of the interface.

What once felt like overhead starts to feel like structure.

What once felt slow starts to feel deliberate.