Security awareness training as a defense against phishing is dead. It has been dead for a while. The industry never held a funeral because the training budget is comfortable, the compliance box gets checked and no CISO wants to tell the board that the program everyone funds does not work.
The premise was simple. With enough education, users would learn to spot the tells. Misspelled words. Awkward phrasing. Sender domains that looked almost right. URLs that revealed something suspicious on hover. We trained a generation of employees to play Where’s Waldo with their inbox, scanning for the one visible artifact that would mark a message as malicious.
Those artifacts are gone. AI-generated attacks are fluent. The infrastructure behind them looks legitimate. The surface signals we trained users to rely on no longer exist. Even if they did, the model would still depend on something humans cannot deliver. Sustained vigilance across hundreds of messages a day, every day, with one lapse leading to compromise. No human attention system works that way.
If user attention is not the answer, what is?
Kahneman applied to organizations, not individuals
Most discussions of phishing lean on author Daniel Kahneman’s System 1 and System 2. Fast thinking is automatic and easy to fool. Slow thinking is deliberate and more accurate. The conclusion is always the same. Train people to slow down.
The framing is true about cognition and incomplete as a security strategy. It asks individuals to sustain behavior that breaks under real conditions.
The more useful application is at the organizational level.
Every company has processes that run fast and processes that run slow. The difference is not accidental. Fast processes are the ones where trust has already been granted and friction has been removed. Wire transfers between known parties. Vendor banking updates. Calendar invites accepted without inspection. Help desk verification over the phone.
Slow processes are the opposite. Trust is being established in real time. Employee logins with conditional access. New vendor onboarding. Any interaction with someone outside the organization.
Most companies did not design this split deliberately. It emerged over time. Someone removed friction because it helped the business move faster. Often, that decision made sense at the time. The threat landscape that justified it no longer exists.
Attackers understand this better than we do. They map where the fast paths are. They wait for moments where scrutiny is minimal. Then they step directly into those lanes.
The Nexus pass as a security primitive
Border control solved a problem that security still struggles with. Uniform scrutiny does not work. Check everyone the same way and movement stops. Check no one and the border disappears.
The solution was risk tiering. Pre-vetted travelers earn a fast lane based on evidence. Everyone else goes through full inspection. The trust is continuously verified and can be revoked the moment new information appears.
The fast lane is not a flaw. The full check is not overkill. Both exist because the system asks the right question. Not whether to trust or verify, but which interactions deserve speed and what evidence supports that decision.
Apply that lens to an enterprise and the gaps become obvious.
Which processes are running on a fast lane that no longer make sense? A vendor whose banking details change over email. A supplier using a typosquatted domain that slips through. A calendar invite from a name that looks familiar enough. An API credential tied to a vendor that has not been active in years.
Each of these is a fast path. Each one has been exploited at scale by attackers who know the assignment was never revisited.
The answer is not to slow everything down. That is the same mistake as awareness training, just applied to processes instead of people. It would destroy productivity and still fail to stop attacks.
The real work is targeted. Identify which fast paths were built on outdated assumptions. Re-tier those. Pull the fast lane from the processes that no longer deserve it. Leave it where it still holds.
The trust inversion no one wants to admit
This leads to a harder question about architecture.
Over the last decade, we applied zero trust to employees and standing trust to suppliers. Employees authenticate constantly. They deal with device checks, session limits and conditional access. Suppliers send a SOC 2 report once and receive long-lived access to critical systems.
That asymmetry deserves scrutiny.
Suppliers are often the path of least resistance for attackers. They hold legitimate credentials. They have access across systems. Many major breaches over the past five years started with a compromised vendor account that was already trusted.
SOC 2 does not solve this. It measures internal control discipline. It answers whether a company follows its processes. It does not tell you whether that company is secure right now.
Yet many organizations treat it as if it does. They make high-stakes access decisions based on a document that was never designed to answer that question.
Compliance automation has made this worse. It turned an annual exercise into a continuous one without changing what is being measured. The bar stayed the same. We just got faster at producing evidence that it was met.
A clean report next to a vendor with an old, compromised credential still active in production is not an edge case. It is a common state.
What deliberate design actually looks like
The work ahead is not glamorous. It will not show up neatly on a dashboard.
Start by mapping processes across the organization. Identify which ones run fast and which run slow. For every fast path, ask three questions.
What evidence originally justified the speed? Does that evidence still hold given current attacker capability? If you remove the fast lane, is the cost lower or higher than the expected impact of a breach tied to that process?
When the evidence no longer holds and the cost of change is lower than the potential loss, the assignment needs to change.
That change will have a cost. Vendor updates that took seconds may take minutes. Help desk interactions may require secondary verification. Onboarding new suppliers may slow down.
The case for accepting that cost is not that caution is good in theory. It is that the original speed was based on assumptions that no longer apply. The efficiency was borrowed from a future failure.
If you cannot explain why a process still deserves a fast lane, you are not making a business decision. You are accepting risk without acknowledging it.
This is what it means to design deliberately. Not forcing everyone to slow down, but making conscious decisions about where speed belongs and where scrutiny is required. Revisiting those decisions as conditions change. Removing fast lane status, the moment it is no longer justified.
This article is published as part of the Foundry Expert Contributor Network.
Want to join?