New · Intent Workflows
Who checked?
It’s the question the board asks after a bad payment, and most CISOs can’t answer it. Detection stops the attacks. For the genuine requests that can’t be undone, Intent Workflows puts your policy in front of each one: alert your team, label it with your rule, route it to an owner, or hold it until a named person has verified it. And you can show it.
- of attacks carry 98% of the modeled financial risk. Phishing Genome
- 3.6%
- of US organizations were affected by business email compromise in 2025. AFP 2026
- 74%
- Deploy time
- 15 min
Brackwell Supply billing@brackwellsupply.com
Updated bank details for invoice 4471
ACME policyBank changes are confirmed by phone to the number on file.
- Vendor history
- Known vendor, 14 invoices paid since 2023
- Authentication
- Passed SPF, DKIM and DMARC
- Payee bank
- New account
- Amount
- $18,400, above your $5,000 threshold
- Owner
- John Doe, AP lead
Released only after John confirms the call. Name and time are kept on record.
Security team notified · Approval request sent in Microsoft Teams
The policy exists. The check often doesn’t.
Every policy says call the vendor. Few can show it happened.
A real supplier really does change banks. The email is genuine, it really is from your vendor, and it still shouldn’t move money without a call to the number you already hold. No filter can make that call. Intent Workflows shows your team every request like it, and where you choose, makes sure someone calls and confirms that they did.
-
Vendor banking change
New bank details mean a call to the number on file.
You choose: hold the email until the payment owner confirms they called the number on file, route it to the payment owner, or let it through with your rule on it and an alert to security. First invoices from new vendors work the same way.
-
Wire transfers
At or above your threshold, a named approver.
You set the amount. Wires at or above it can wait for the owner you choose, who signs in through your identity provider before they can release. Below it, security still sees every one.
-
Payroll and VIP access
Routed to the right team before anyone acts.
Payroll deposit changes can be verified the same way. Password, MFA and account recovery resets for your executives, including those sent to the help desk, go to your identity or security team, as an alert or a hold.
-
Policy banners
Your rule, at the moment it matters.
Banners appear only on the requests you track and state your own policy, such as “Wire transfers must follow ACME policy. Contact accounts@acme.com.”
Intent for good email
Most email security reads intent only once an email is already bad.
That leaves the most dangerous message of all, a real vendor changing where money goes, delivered like any other. StrongestLayer reads what every email is asking someone to do, including the genuine ones. That is what lets detection and your policy do two different jobs on the same request.
- Quarantined
Compromised vendor sends new bank details
Detection sees the takeover. Nobody ever receives it.
- Held for verification
Real vendor sends new bank details
Genuine email. Still not acted on until a named person confirms the callback.
You set the rules
Alert, label, route or hold each kind of request.
Nothing is on until you turn it on. For each request type, you decide how far to go: alert your security team and simply see every one, deliver it with your policy banner, route it to a process owner, or hold it for out-of-band verification. A workflow can be visibility only. Many teams start in alert mode for a month, then tighten the ones that matter. You set the wire threshold, from every payment to those at or above, say, $5,000, and how long a release link stays live.
- Microsoft 365
- Google Workspace
- Microsoft Teams
- Approvals arrive in Microsoft Teams or email
- Approvers sign in through your own identity provider before they see a held email
- Every hold ends with a name, a time and a record
- Works alongside your existing email security
Also new: AI Instruction Intercept
Stop hidden instructions before your AI assistant reads them.
The newest board question is whether an email can turn your AI assistant against you. In 2025, a Microsoft 365 Copilot flaw (CVE-2025-32711) showed a single crafted email could make the assistant disclose data. The same engine that reads what an email asks a person to do also reads what it asks a machine to do. AI Instruction Intercept uses a reasoning model and your company’s own context to stop hidden instructions before Copilot, Google Gemini or any AI that reads your mail sees them.
-
Quarantined by default
The hidden instruction never reaches the assistant.
Email carrying instructions written for an AI is quarantined unless you choose otherwise.
-
Explained
Each instruction’s intent and target, in plain words.
Your team sees what the hidden text was trying to make the assistant do, and what it targeted.
From our chief product officer
“Prompt injection can reach an AI assistant many ways, but email is the one any outsider can use. Planting an instruction on someone’s desktop takes real access. Sending them an email takes an address. That makes the inbox the first place to stop it.”
FAQ
The questions the board asks.
What is Intent Workflows?
Intent Workflows is a StrongestLayer capability that reads what an email is asking someone to do and applies your policy to the requests that can’t be undone: vendor banking changes, wire transfers, first-time vendor invoices, payroll deposit changes and executive credential resets. For each one you choose the action: alert your team, deliver it with a policy banner, route it to the right owner, or hold it for out-of-band verification. It is available now for Microsoft 365 and Google Workspace.
How do we protect ourselves from vendor email compromise?
In layers, with a person in the loop. Detection stops the compromised vendor emails before anyone receives them. Any genuine request to change where money goes, or to pay a vendor for the first time, is held until the owner you named has called the number on file and confirmed it. Every one of those checks is on record.
How is this different from detection?
Detection decides whether an email is malicious. Intent Workflows decides whether your policy requires a check before the action, whatever the verdict. A real vendor changing banks is a perfectly good email, and your policy still says call them first. Detection can’t make that call. This makes sure someone does.
How do we know the verification actually happened?
The email is not released until the condition you set is met: the call confirmed by a named person, the approver signed in, the owner’s sign-off given. Each hold ends with a name, a time and what they did, so the answer to “who checked?” is a record rather than a policy document.
Do we have to hold email?
No. Each workflow can alert, label, route or hold, and you choose per request type. Some teams hold every bank change from day one. Others start by seeing every sensitive request and adding holds where the risk justifies the friction. Either way, security has visibility on all of them.
Who can release a held email?
Only the owner your policy names. Approvers sign in through your own identity provider, Google or Microsoft, before they can see a held email, with an optional second step for high-risk requests. Release links expire on the schedule you set.
What is AI Instruction Intercept?
AI Instruction Intercept quarantines email that carries instructions written for an AI assistant before Copilot, Gemini or any AI that reads your mail sees them, and explains in plain words what each instruction was trying to make the assistant do and what it targeted. It launches alongside Intent Workflows and is on by default.
Do you have an AI gap? Find out what’s getting through.
See which generations of defense you’re running and what it takes to close the gap. Two minutes. No signup. Nothing changes in your environment.