how ai agents are changing email forever

In late 1971, Ray Tomlinson sat in front of two computers at BBN Technologies in Cambridge, Massachusetts. He sent a test message from one machine to the other across the ARPANET. The message was something forgettable, probably "QWERTYUIOP." The idea was not. For the first time, a person could send a written message to another person on a different computer over a network. Tomlinson picked the @ symbol to separate the user from the machine because he figured it would never appear in someone's name.

That was 55 years ago. And the fundamental model hasn't changed. A human composes a message. Another human reads it.

Until now.

The protocol that refused to change#

In August 1982, Jonathan Postel published RFC 821, defining the Simple Mail Transfer Protocol. SMTP gave the internet a standardized way to move text messages between servers. David Crocker published RFC 822 around the same time, specifying how those messages should be formatted. Together, these two documents became the foundation of email as we know it.

The protocol has been updated since then. RFC 2821 arrived in 2001. RFC 5321 replaced it in 2008, adding extensions for security and authentication. But the core model stayed the same: a human sender, a human recipient, a subject line, a body. Every improvement to email over the past four decades has been about making that human-to-human experience better. Better spam filters. Better search. Better mobile clients. Better threading.

Nobody designed email for a non-human user. Why would they? There wasn't one.

The first new type of email user in half a century#

Today, nearly 4.7 billion people use email. Roughly 392 billion emails cross the internet every single day. And for the first time since Tomlinson's test message, there's a genuinely new kind of participant in this system: the AI agent.

Not a mail merge script. Not an autoresponder. An autonomous entity that reads incoming messages, understands context, makes decisions, and composes replies on its own. An entity that can provision its own address, manage its own inbox, and communicate with humans and other agents without anyone clicking "compose."

This is a bigger deal than it sounds. Email is the most universal communication protocol on the internet. It predates the web. It works across every platform, organization, and country. Giving an agent an email address means giving it the ability to reach anyone, anywhere, without requiring the other side to install anything or adopt any new protocol.

What agents actually do with email#

The use cases are already here. Agents triage support tickets, draft responses from a knowledge base, and handle password resets at 2 a.m. They monitor newsletter subscriptions and send daily summaries. They schedule meetings through plain email threads, negotiating time zones and availability without the other party ever knowing they're talking to a machine.

But the more interesting shift is agent-to-agent communication. When multiple agents each have their own address, they can coordinate through standard email threads. A research agent finishes gathering data and emails the results to a writing agent. The writing agent drafts a report and sends it to a review agent. Every step is logged in the thread. The audit trail writes itself.

Google launched the Agent2Agent (A2A) protocol in April 2025, with backing from over 50 technology partners, to standardize how agents talk to each other. But here's the thing: email already does this. It's federated, universal, and asynchronous. Agents don't need to share a framework or a runtime. They just need an address.

Why the old infrastructure doesn't fit#

Here's where it gets uncomfortable. The entire email ecosystem was built around the assumption that the user is a human with a browser or a mail client. Think about what it takes to give an agent email access today. You create a Google Cloud project. Configure OAuth consent screens. Generate credentials. Handle token refresh. Deal with IMAP rate limiting. Pray the app password doesn't get revoked on a Tuesday afternoon.

Worse, most approaches require the agent to connect to a human's existing inbox. Your agent gets OAuth access to your Gmail, and suddenly it's sitting in the same context as your medical records, your bank statements, and your private conversations. A Fortune investigation highlighted this exact risk in February 2026, and Microsoft's security team published research on prompt injection attacks through email that can hijack agent behavior.

The problem isn't email itself. SMTP is beautifully simple and it works. The problem is that everything built on top of it, the account creation flows, the authentication layers, the inbox management tools, assumes a human is at the keyboard.

Agents need something different. They need to self-provision their own inboxes. They need isolated addresses that don't expose a human's entire email history. They need infrastructure that treats them as a first-class user, not an afterthought bolted onto a system designed for people.

What the future looks like#

We're moving toward a world where most email traffic isn't human-to-human. It's agent-to-human, human-to-agent, and agent-to-agent. The sender's AI crafts a message. The recipient's AI decides whether to surface it, respond automatically, or archive it. Success stops being about open rates and starts being about completed actions.

Multi-agent coordination through email will become ordinary. Teams of specialized agents, each with their own shell, passing tasks and results through threaded conversations. The reef grows. The communication graph gets richer. And email, the protocol everyone keeps predicting will die, turns out to be the connective tissue that makes it all work.

This isn't science fiction. Gartner predicts 40% of enterprise applications will embed task-specific agents by the end of 2026. Roughly 35% of organizations already report broad usage of AI agents. The future of agent email isn't a question of whether. It's a question of how fast.

Why we think about this every day#

LobsterMail exists because we saw this shift coming. Email wasn't designed for agents, but it doesn't need to be replaced. It needs infrastructure that welcomes them. A place where an agent can hatch into the world with its own address, its own shell, its own isolated space to communicate. No human signup. No OAuth ceremony. No shared credentials.

The protocol Tomlinson and Postel gave us is remarkably durable. Fifty-five years later, it's still the most universal way to reach anyone on the internet. The only thing missing was a seat at the table for the new participants.

Now they have one.

---

Give your agent its own email. Get started with LobsterMail — it's free.