Superhuman pushes AI email from assistance toward automation

Superhuman is moving AI email from a tool users summon into a system that prepares work in advance.

What happened

The company launched an updated auto-drafting feature that identifies incoming messages likely to require a response and automatically prepares several possible replies. The drafts are based on a user’s previous conversations, writing style and the context of the email thread.

Users can select, edit or reject the suggestions, and the system is designed to learn from those decisions over time. This is a more proactive model than conventional writing assistants, which generally wait for a user to open an email and request help.

The product still has clear limits. An AI-generated reply can sound appropriate while making the wrong commitment, accepting an unsuitable meeting or responding with more enthusiasm than the user intended. The final decision therefore remains with the person sending the message.

Why it matters

Email is a high-frequency workflow where even small time savings can compound. The commercial value is not simply producing nicer sentences; it is reducing the cognitive cost of deciding which messages need attention and drafting routine responses.

Superhuman also has an advantage over general-purpose writing tools because it operates inside the inbox and can use thread history, contact context and previous behaviour. That integration can make the output more personalised, but it also raises expectations around privacy and data handling.

The bigger picture

Enterprise software is shifting from copilots that answer questions toward agents that monitor work and prepare actions. The useful boundary is likely to be “automate the preparation, preserve human approval” for tasks where mistakes have reputational or contractual consequences.

Email may become an early test of whether proactive AI genuinely reduces workload or simply creates another queue of machine-generated suggestions to review.