Powered byAnthropic

Phishing SOC Agent

Any incident gets an L2-level investigation in minutes, with every step on the record.

Watch the Phishing SOC Agent work one incident from question to verdict.

How it works

It finds the one thread in each email worth pulling and follows it to a verdict.

Every investigation opens with the checks any analyst would run. Then your Phishing SOC Agent looks for what makes this email different, and each answer it finds decides the next question.

Illustrative examples based on real investigations. Brands, names, prices, numbers and domains are fictional.

Where teams use it

Your Phishing SOC Agent steps in when a person on your team has to make the call.

Adaptive AI already handled these emails. Each story starts with a question someone asked about one of them and ends with an answer your team can stand behind.

1Verifying a closed incident

Already handled. Now explained.

The moment

Your CFO asks why an email from a board member never reached her inbox. Adaptive AI quarantined it last night.

The agent

One click hands the incident to your Phishing SOC Agent. It plans the checks, then sends sub-agents to look at the sender, the links, the attachment and the route the email took.

The outcome

In minutes you have a Phishing verdict and the evidence behind it. You forward the case file to the CFO, and the question is closed.

Incident · Quarantined#48213
From
James Ramsey <d.ramsey@northfield-capita1.com>
To
Office of the CFO
Subject
Board pack for Thursday, please review tonight
Closed by Adaptive AIQuarantined
Lead investigator: checking sender, links, attachment, delivery path
  • SenderLookalike of the board member's domain, registered 9 days agoSuspicious
  • Links"Open board pack" redirects twice to a credential pageMalicious
  • AttachmentAgenda PDF opened in isolation, no active contentClean
  • DeliverySent through a bulk relay the real sender has never usedMismatch
Phishing
Key evidence: lookalike domain, credential-page redirect
Every step logged
2Ironscales Autopilot check-ins

Trust Ironscales Autopilot. Check in on it whenever you want.

The moment

Ironscales Autopilot resolves incidents on your behalf, end to end. It handled 1,618 this month, your quarterly review is next week, and you want a few documented examples.

The agent

Autopilot never calls the agent on its own. You pick an incident from the list, and the agent works it the way a skilled L2 analyst would, recording every search and lookup as it goes.

The outcome

You get the full story behind what Autopilot did, in a form you can drop straight into the review deck.

Autopilot · Resolved this month1,618 incidents
  • "Updated W-9 attached" · accounts payableRemediated
  • "Your parcel is held at customs" · 14 mailboxesRemediated
  • "Shared: Q3 bonus schedule" · HR teamRemediated
  • "Voicemail from +1 (614) 555-0142" · 3 mailboxesRemediated
Lead investigator: checking sender, message, attachment, delivery path
  • SenderFirst email ever from this domain to your organizationNew sender
  • MessageAsks AP to change a vendor's bank details before FridayPayment fraud
  • AttachmentW-9 opened in isolation, bank details differ from the vendor recordMismatch
Phishing
The full why behind what Autopilot did
Report ready to share
3Release requests

Say yes or no to a release with the evidence in hand.

The moment

An employee in facilities asks you to release a quarantined invoice. The vendor says payment is overdue.

The agent

The agent checks who really sent it, where every link goes and what the attachment contains.

The outcome

The verdict comes back Safe, with the evidence attached. You release the email and keep the case on file for anyone who asks later.

Release request · Pending your answer#48977
From
Billing <invoices@harborline-hvac.com>
Requested by
Facilities coordinator
Subject
Invoice 2291, 30 days past due
"Can you release this? They say we're late."Quarantined
Lead investigator: checking sender, links, attachment
  • SenderHas written to your organization for two years, authentication passesKnown
  • Links"Pay invoice" goes to the vendor's real billing portalLegitimate
  • AttachmentInvoice PDF opened in isolation, no active contentClean
Safe
Key evidence: long sender history, real vendor portal
Released by you

Illustrative examples. Names, domains and figures are fictional. The agent reports a verdict, and your team makes every decision.

Use cases

Three moments when your team owns the decision.

Verifying closed incidents

Already handled. Now explained.

A VIP sender, an odd campaign or a question from leadership. You get the full why behind the outcome, ready to share.

Ironscales Autopilot check-ins

Trust Ironscales Autopilot. Check in on it whenever you want.

You want a periodic look at what Ironscales Autopilot handled on your behalf, and the agent runs only when you ask. You get a fast way to spot-check incidents and document what you found.

Release requests

Say yes or no to a release with the evidence in hand.

An employee asks for a quarantined email back. You get a documented case for your answer, for the user and for anyone who asks later.

Every step on the record

Every search and every finding goes on the record.

You see each search, lookup and finding as it happens, then a verdict with the key evidence behind it. Use the case file for escalations, user follow-ups, leadership reports and compliance reviews.

Under your control

Your team stays in charge of every decision.

An AI investigator should be clear about what it does and what it leaves to you.

You start every investigation.The agent runs only when an admin opens an incident and clicks Generate.
It never acts on its own.The agent reports a verdict. It never reclassifies, releases or remediates anything.
Every step is on the record.Each search, lookup and finding is logged, then summarized with the verdict.
Attachments open in isolation.The agent opens files in an isolated sandbox.
You decide who uses it.Admins control who in your organization can use each agent.
How it fits

Adaptive AI handles the volume, and the agent takes the closer look.

Adaptive AI protects every mailbox automatically, and your thresholds decide when it acts on its own and when it brings an incident to you. When you want to look closer, the Phishing SOC Agent does the legwork.

AttributeAdaptive AIPhishing SOC Agent
JobDetect and remediate threats across every mailboxInvestigate one incident in depth
When it runsAlways, automaticallyWhen an admin asks
SpeedMachine speed, at scaleMinutes per incident
OutputA classification and automatic actionA verdict plus the full evidence trail
Who it servesThe whole organizationThe analyst making a call

Adaptive AI keeps protecting every mailbox whether or not anyone runs the agent.

FAQ

Questions security teams ask about the Phishing SOC Agent.

What is the Phishing SOC Agent?

The Phishing SOC Agent is an on-demand AI investigator inside Ironscales. It investigates any email incident you hand it the way a skilled L2 analyst would, then delivers a verdict of Safe, Phishing or Spam with the key evidence and a full record of every step.

If Adaptive AI already protects us, why do we need an investigator?

Adaptive AI protects you whether anyone is watching or not. Some decisions still belong to a person: releasing a quarantined email, signing off on a borderline incident, explaining a call to an executive or an auditor. The agent gives that person the evidence to decide fast and defend the decision later.

Does the agent run on every email?

No. An admin starts every investigation. Most incidents don't need a case file because Adaptive AI already handled them. The agent is there for the ones your team wants to understand, prove or explain.

Can the agent change or release an email on its own?

No. The agent reports a verdict and never reclassifies, releases or remediates anything. Your team reads the case, makes the decision and moves on to the next one.

What does the agent check?

A lead investigator plans each investigation and sends sub-agents to check the sender, the message, every link, every attachment in an isolated sandbox, the delivery path, and your organization's context, such as who else received the email and the recipient's role.

How do I start an investigation?

Open any incident in your Ironscales console, select the Phishing SOC Agent tab and click Generate. In a few minutes you get a verdict, the key evidence behind it and a full record of every step the agent took.

Phishing SOC Agent

Your Phishing SOC Agent runs the deep investigation, and you see every step of it.

You get forensic depth when you need it, and your team stays focused on everything else.