Phish your own people before a real attacker does.
An AI agent researches each person and writes a lure just for them, the way attackers now work. Your security team scopes and approves every run.
security teams at:
Phishing tests teach people to spot the template.
Real attackers don’t reuse templates. They research one person and write one email. A library lure measures recognition, not exposure.
click-through on fully AI-written spear phishing, the same as human experts. Generic phishing got 12%.
Heiding et al., 2024, 101 participantsIllustrative example. Names and domains are fictional.
Why the test has to be this good.
Watch one engagement run. Your team approves every step.
The agent scopes, researches, writes, sends and measures, then hands every lure and source back to your security team.
- 01Scope
- 02Research
- 03Craft
- 04Review
- 05Measure
- 06Report
Set the scope
Nothing runs until your security team approves it.
- Targets
- Finance · 14 people
- Timing
- Tuesday to Thursday, working hours
- Tone
- Vendor and colleague impersonation
- Credentials
- Never captured or stored
Research each person
Public sources only: web, LinkedIn, news and code search. 3 of 14 shown.
| Person | What the agent found | Sources |
|---|---|---|
| Dana R.Accounts payable | Runs vendor payments. Named Corvane Supply in a case study. | 3 |
| Omar K.Payments systems | Maintains the invoicing tool. Mentions it in a public repo. | 2 |
| Priya S.Finance operations | Spoke about quarter-end close at a conference. | 4 |
Write a lure for each person
Two to three per person, reviewed by your team before anything sends.
Hi Dana, following our Q4 remittance call, payments now go through our new supplier portal. Please log in before Friday so the next run isn't delayed.
Your team reviews every lure
Nothing sends until the security team has read and approved it.
- Lures written
- 38 for 14 people
- Read by
- Security team
- Sender
- Approved sender, on the agreed schedule
- Stop
- Available at any time
Measure what each person did
Every event timestamped, per person. No password is ever kept. 4 of 14 shown.
| Person | Lure | Event |
|---|---|---|
| Dana R. | Vendor remittance | Submitted 09:43Clicked 09:42 · no password kept |
| Leo M. | Colleague file share | Submitted 11:18no password kept |
| Omar K. | Tool access share | Replied 10:05 |
| Priya S. | Colleague budget review | No action |
Evidence for your team
Handed back with every lure and the sources behind it.
- TrainingVendor payment changes for accounts payable
- ControlsPortal links from new vendor domains
- InventoryEvery lure, source and event, ready to export
Three ways the agent gets in. Pick one to see the lure.
Following our Q4 remittance call, payments now go through our new supplier portal. Please log in before Friday.
- SourcesA company case study naming Corvane Supply; her role on LinkedIn
- Why it worksA real vendor and a real process, with one new link
- ResultClicked 09:42, submitted 09:43. No password kept
A teammate shared the invoicing workspace with you. Reply to confirm access before the change freeze.
- SourcesA public code repository that names the tool; his team page
- Why it worksA tool he uses every day and a routine request
- ResultReplied 10:05 asking for access
Can you look at the updated close budget before our 3pm review? Sending from my personal inbox, laptop's being fixed.
- SourcesA conference talk on quarter-end close; the company team page
- Why it worksA real colleague and a real meeting
- ResultNo action taken
Behind the console: seven steps, one approved scope
Your security team approves every step and can stop a run at any time.
An offensive capability, run safely.
- Consent first.Every run is scoped and approved with your security team before anything sends.
- Your people, your terms.You define targets, timing and tone, and you can pause or stop a run at any time.
- An auditable trail.Every lure, every source and every event is logged and handed back to you as evidence.
What your security team will ask
Is this safe to run against employees?
Yes. Every run is scoped and approved with your security team first. You set targets, timing and tone, and can stop it at any time.
How is this different from phishing simulation?
Simulation replays attacks Aegis caught in your tenant. The red team agent researches each person and writes a new lure, the way an attacker would.
Are real credentials captured?
No. The agent records that a submit happened on the lure page. Passwords and secrets are never collected.
What happens to the research?
It comes from public sources and is handed back to you in the report, with every lure.
Who can be targeted?
Only the domain, teams, roles or named people you approve.
See who clicks before an attacker does.
Scope a run with your security team and get per-person evidence back.
See what Aegis finds in your environment.
Connect a tenant, run in monitoring mode, and compare against what your current stack delivered.