Clear the innocent alerts with evidence.
Most takeover alerts have an innocent explanation, like a trip, a new laptop or a late night. Agentic ATO proves it and shows why, so your analyst confirms the call instead of rebuilding the investigation.
Learn how global companies are transforming their email security to tackle modern threats
Learn how MSPs can revolutionize their email security and profitably grow their businesses
Secure email gateways and API-based solutions compared for enterprises and MSPs.
Get an overview of our API, Adaptive AI, and Human Element components
Explore the benefits of a mailbox-level, API-driven email security platform
Learn how our Adaptive AI blocks attacks missed by other solutions
See how continuous human insights maximize defense capabilities
Our agents work in concert to anticipate attacks, investigate threats, and educate users
Researches your org like an attacker would, then blocks those threats before they arrive
L2-level forensic investigation across five tracks. Clear verdict in minutes.
Hyper-personalized simulations targeting your highest-risk employees with real OSINT
Stop attacks like BEC, VEC, and VIP impersonation
Continuously protect against malicious links and attachments
Prevent, detect, and respond to ATO attacks in real time
Encrypt outbound email and meet compliance, automatically
Agentic ATO investigates every account takeover alert the way an analyst would, then hands back a verdict with the evidence and its reasoning.
Sign-in from Lisbon matches a trip on the calendar.
Illustrative example.
If you can draw a flowchart of how a tool works, it's automation. Agentic ATO picks the checks each case needs, reads the results and follows the evidence where it leads.
Someone signs in to your controller's mailbox from Lisbon. Your account takeover detection flags it, and now someone has to work out whether that's an attacker or a controller at a conference.
Agentic ATO reads the alert and writes its own plan. A sign-in from Lisbon and a 2 a.m. sign-in from a new laptop raise different questions, so they get different investigations.
The agent checks the topics it chose and marks each one Suspicious, Looks normal or Inconclusive. Sent items and Travel & calendar appear on every incident.
Every verdict comes back Safe or Compromised, with a confidence score, the evidence and a plain-language explanation your analyst can check in a minute.
In auto mode, a Compromised verdict blocks the user and closes all active sessions, and a Safe verdict classifies the incident. Change the facts above to see both outcomes.
Sign-in from Lisbon, Portugal at 6:12 a.m.
The open question. Is this an attacker, or a controller who flew to Lisbon for a conference?
The sign-in lines up with a flight and a hotel booking on the user's calendar. The device, mailbox rules and sent items all match how this user normally works.
There is no trip on the calendar, the device is new, a rule now forwards mail outside the company, and the account sent a burst of messages to finance right after the sign-in.
Your analyst can reverse any call. Marking a case Safe can unblock the user in the same save.
Illustrative example. The user, domain and incident are fictional.
Most takeover alerts have an innocent explanation, like a trip, a new laptop or a late night. Agentic ATO proves it and shows why, so your analyst confirms the call instead of rebuilding the investigation.
A new forwarding rule and a burst of mail to finance follow an odd sign-in. In auto mode, the agent blocks the user and closes all active sessions, with the findings on the incident.
Each incident carries the verdict, a confidence score, the topics the agent chose and its reasoning. Your analyst reviews it quickly, and the record stays on the incident.
Auto mode is on by default, and every setting stays in your hands under Settings, Account takeover.
Detection works the way it does today. Agentic ATO adds the investigation that used to wait for an analyst, and Adaptive AI keeps protecting every mailbox the way it always has.
| Attribute | Account takeover detection | Agentic ATO |
|---|---|---|
| Job | Flag sign-ins and account behavior that look unusual | Prove whether the account is compromised |
| Output | An alert | A verdict of Safe or Compromised, with a confidence score, the evidence and the reasoning |
| Next step | An analyst investigates | Auto mode acts, or Suggest mode recommends and your analyst decides |
| Sensitivity | Relaxed, Balanced or Aggressive | The same setting shapes how the agent behaves |
Agentic ATO works with Microsoft 365 for customers with Ironscales Account Takeover protection, with nothing to install.
Agentic ATO is a capability in Ironscales Account Takeover protection. It investigates each account takeover alert the way an analyst would, then returns a verdict of Safe or Compromised with a confidence score, the evidence and its reasoning in plain language.
It decides its own steps. Agentic ATO reads the alert, picks the checks the case needs, reads the results and follows the evidence. Sent items and Travel & calendar appear on every incident, and the rest is the agent's call.
In auto mode, which is on by default, a Compromised verdict blocks the user and closes all active sessions, and a Safe verdict classifies the incident. You choose the actions, and your analyst can reverse any call. Turn auto mode off for Suggest mode, where your analyst makes every final call.
It chooses from sign-in history, devices, mailbox rules, sent items, the calendar and how peers behave, based on what each alert needs to answer.
Analysts mark a case Safe or Compromised, or give it a thumbs up or down with an optional comment. Those classifications and that feedback go to the agent for your tenant.
Agentic ATO works with Microsoft 365.
On the Alerts & notifications page, you choose who gets an email and for which agent events: auto-resolved Compromised, auto-resolved Safe, or Case needs my review in Suggest mode.
Your analysts get a verdict, the evidence and the reasoning, and they keep the final word on every call.