Outlook and Microsoft 365 for AI agents: what works and what breaks

Use Outlook and Microsoft 365 when the mail belongs to a human or a team and an assistant helps inside it, under your admin's controls. Use an inbox built for agents when the agent reads and replies alone, hears from strangers, or needs its own address and codes, screened before delivery.

Updated

Outlook on Microsoft 365 is an excellent mailbox for a human: anti-malware, anti-spam and anti-phishing protection is on by default for all cloud mailboxes, with strong deliverability and admin tools. An AI agent is a different reader.

Outlook and Trotta Mail side by side

Outlook and Microsoft 365Trotta Mail
Built forA human or a team reading their own mail.AI agents and humans. The guard is stricter for an agent.
When mail is screenedBefore delivery, by connection, malware, spam and phishing filters. Zero-hour auto purge acts on mail already delivered.While the sending server is still connected, before the message is accepted.
What an agent can readMail.Read reads a mailbox. As an application permission, all mailboxes until an admin scopes it.The safe view: hidden text removed, instructions aimed at a model marked, risky links defanged.
Held mail and human reviewMalware and high confidence phishing go to quarantine, for admins by default. Spam goes to Junk Email.Held outside the inbox. The agent sees a verdict, no content. A human can release it.
Outgoing checksOutbound spam filtering and controls on automatic external forwarding. Data loss prevention policies with Microsoft Purview, depending on license.Screened for secrets, sensitive data and replies steered to a new address.
Sign-in codesReadable with Mail.Read.Sealed from every API key in a human's mailbox. Parsed for the agent in its own inbox.
Setup for many agentsA mailbox each, a service principal, and an admin to scope its permissions in PowerShell.One API call per inbox. Pods keep each customer's inboxes apart.
Price to startA license for each user mailbox. A shared mailbox needs none up to 50 GB.Free: 2 inboxes, 1,000 emails a month. Developer, $20 a month: 5 inboxes, 5,000 emails. Startup, $200 a month: 50 inboxes, 50,000 emails.

What works

  • Screening before delivery. Connection filtering rejects most spam, malware goes to quarantine, and high confidence phishing is always quarantined. A message that clears these layers is delivered.
  • Deeper checks with Defender for Office 365. Safe Attachments opens attachments in a virtual environment before delivery, and Safe Links checks a link again when it is clicked.
  • A supported API. Microsoft Graph reaches primary and shared mailboxes.
  • Mailboxes for agents, in preview. Agent 365 documents an agent user account that can hold a mailbox, for tenants in the Frontier preview program.

What breaks when an agent reads the mailbox

  • A real-looking message can get through. Microsoft's docs say malicious messages can still be received, for example when content is weaponized after delivery. Zero-hour auto purge then acts on malware and phishing already in the mailbox, within the last 48 hours of delivered email. An agent may have acted by then.
  • The agent reads everything. As an application permission, Mail.Read reads email in all mailboxes without a signed-in user, until an admin scopes it. Sign-in codes and reset links are messages. The narrower Mail.ReadBasic leaves out every body.
  • One crafted email was shown to be enough. In EchoLeak, CVE-2025-32711, researchers showed that one crafted email could make Microsoft 365 Copilot disclose data with no click. Microsoft rated it 9.3 (critical) and published its advisory in June 2025, saying the flaw was already mitigated. Microsoft lists it as not exploited.
  • A shared sign-in is out. Microsoft removed Basic authentication for IMAP, POP and Exchange Web Services, so apps use OAuth. A shared mailbox is not intended for direct sign-in. For personal Outlook.com accounts, the Microsoft Services Agreement says account credentials cannot be transferred.
  • Many agents is heavy setup. With application permissions, each agent needs a mailbox, a service principal and an admin to scope its permissions. A change notification subscription for mail lasts under seven days, then must be renewed.

When Outlook is the better choice

  • The mail is a human's or a team's own, and an assistant helps inside it.
  • Your organization runs on Microsoft 365 and wants agents under the same admin controls.
  • A team already works a shared mailbox, and the agent assists its members.
  • You need Outlook, IMAP or a desktop mail client. Trotta Mail has no IMAP or SMTP access for mail clients.

When a pre-delivery inbox is the better choice

  • The agent reads, decides and replies with no human checking each message.
  • The agent signs up for services and needs its own verification codes, apart from a human's.
  • Strangers can write to the agent, so mail must be checked for text written to hijack a model.
  • You need many inboxes, created by code.

No filter catches everything, Trotta Mail included. A reworded scam with no technical tell can still be delivered, which is why uncertain mail is held for a human.

What moving looks like

Keep human mail in Microsoft 365 and give each agent an address of its own.

  1. Create the inbox with one call, POST /v1/agents/signup. It returns an address and an API key.
  2. Connect the agent by REST API, MCP server, or the TypeScript or Python SDK.
  3. Change the address on the agent's accounts, then remove its Graph permissions.
  4. On a paid plan, connect a subdomain such as agents.example.com, so mail for your main domain stays in Microsoft 365.

Questions

Can an AI agent use an Outlook mailbox?

Yes, through Microsoft Graph with OAuth. Decide what the agent may read, and who reviews a message that looks real.

Can I use a shared mailbox for an AI agent?

Graph supports shared mailboxes, so it can work. Microsoft says a shared mailbox is not for direct sign-in and supports at most 25 members. The agent reads what every member reads.

Does Microsoft 365 screen email before delivery?

Yes. Built-in filters reject, quarantine or junk spam, malware and phishing before delivery. Zero-hour auto purge acts after delivery, on messages identified later.

Sources

Every check is on, on every plan. Free for two inboxes. No card.