Security architecture is not a document. It's not a certificate. It's the set of decisions that determine whether your technology can withstand the threats it faces. Getting those decisions right is expensive in one way or another — and most companies misunderstand which way.
When people hear "security architecture," they often think of one of two things: a compliance document, or a firewall configuration. Both are wrong. Security architecture is the set of decisions across your entire technology stack that determine your attack surface:
No single person covers all of that. A security architect who knows IAM deeply may not know SIEM detection rules. One who understands encryption may not understand supply chain. The job is actually four or five jobs.
A security architect in the Netherlands costs €80K-120K+/year in salary. Before benefits, tools, and the 6-12 months it takes to find the right person. For most companies, that's the most expensive single hire on the payroll.
And you need more than one skill set. NIS2 wants risk management. DORA wants operational resilience. The AI Act wants governance architecture. GDPR wants data protection by design. ISO 27001 wants 93 controls mapped to technical verification. No single hire covers all of that. You'd need a team.
Consultants produce reports. Architecture diagrams. Gap analyses. Maybe a remediation roadmap. Then they hand you a PDF and leave.
The report doesn't build anything. It describes what someone else should build. And the problem is what happens after: a report doesn't implement itself. ADRs don't deploy themselves. A compliance roadmap doesn't produce evidence for regulators. You're left with a document and a to-do list.
The real cost question isn't "hire or consult?" — it's "how do you get architecture that's technically verified, not just documented?" A report describes security. An architect designs it. Neither verifies it works. Under NIS2 and DORA, regulators want the verification.
NIS2 doesn't say "have a security architect." It says implement risk management measures, handle incidents, secure your supply chain, and report within deadlines. That's architecture, not a job title. Specifically:
DORA (Digital Operational Resilience Act) goes further than NIS2. It wants operational resilience demonstrated through testing. Not a plan that says "we can recover." Evidence that you tested recovery. Not a policy that says "we have threat detection." Evidence that detection was tested against actual attack scenarios.
If you're deploying AI systems, the AI Act requires governance architecture: risk classification, human oversight design, data quality controls, transparency mechanisms. That's architecture, not a compliance checklist. And high-risk AI systems need it built in before deployment, not retrofitted after.
Here's the pattern we see: a company has ISO 27001 on paper. They have a security policy. They may even have a security architect. But the IAM permissions don't match the access control policy. The SIEM rules were never tested against real attacks. The encryption keys are stored alongside the data. The supplier security assessments are collected PDFs, not technical verifications.
The architecture exists on paper. It doesn't exist in the technology. And that's the gap that breaches go through.
The cost question is really: how do you get architecture that's technically verified? Three paths exist:
The third option is what we do at Aethel Guard. Architecture sprints that include security architecture as standard. But the point of this piece isn't to sell — it's to explain the actual cost. The cost of getting architecture right is always less than the cost of getting it wrong.
Aethel Guard is a Netherlands-based security architecture firm. We design and analyse security architecture for projects and products. Your architect, not your vendor.