Zoho has opened early access to AgentInbox, email infrastructure built specifically for AI agents rather than for people. Each agent provisioned through it gets a permanent address, its own isolated credentials, and a complete audit trail of every message it sends or receives, instead of borrowing a shared inbox or a forwarded human alias the way most AI outreach and support agents work today.
Why it matters: outbound sales agents that send email under a shared or borrowed mailbox inherit that inbox’s entire history, recovery links, and legal accountability, and there is no clean way to tell which messages a human sent versus which an agent sent. Revoking access when something goes wrong is messy, and the inbox often doubles as the recovery point for several other critical accounts the agent now effectively has the keys to. AgentInbox handles SPF and DKIM authentication and isolates sending domains per agent, so one compromised or misbehaving AI sender does not damage the deliverability reputation of every other outbound program riding on the same domain.
The original insight: as AI sales agents keep getting assigned actual job titles rather than staying single-purpose prompts, the constraint is shifting from “can the agent write a good email” to “can the company prove, after the fact, which email came from a person and which came from software.” A dedicated, revocable identity per agent is what makes that provable, and it is the same audit and access-control problem that has already pushed sales teams to demand accountability from AI tools marketed as coworkers rather than take the vendor’s vocabulary at face value.
Source: Zoho
