Every agency puts "AI" on the homepage these days. What that concretely means usually stays vague — a tool here, an experiment there. Our answer is a system we can simply show you. This piece is that tour: this is how every change in our projects gets built, checked and approved.
The fair question about a one-person agency
Webmakkers is deliberately one senior builder: no account manager in between, no junior doing the real work. The fair follow-up question is: who checks the work? In a team of ten, a security specialist looks at your forms, a tester hunts edge cases and someone guards accessibility. We didn't cut that scrutiny — we automated it and wrote it down, so it happens every single time, not just when there's time for it.
What's under the hood
Our build system has three layers:
- 85+ specialist AI reviewers. Not one general assistant, but separate reviewers per discipline — security, privacy & GDPR, performance, accessibility, payments, test quality. Every change is judged by the specialists who own that subject, exactly how you'd organise it in a large team.
- 100+ checklists with cited sources. The reviewers don't judge on gut feeling but against written criteria, with sources — from OWASP security rules to GDPR requirements. A checklist doesn't get tired and doesn't skip a step because it's Friday afternoon.
- Automated gates before every release. Code does not leave our system without a row of automated checks: a scan for leaked keys and passwords, static security analysis, an audit of all external dependencies for known vulnerabilities, type checking and code quality. If a check fails, it gets fixed before anything ships.
How a change goes live here
- Build. AI does the typing — under precise instructions, within architecture rules locked down per project.
- Review. The system looks at what was touched and selects the relevant specialists. If a change touches a contact form, the security and privacy reviewers weigh in; if it touches the interface, accessibility and performance do.
- Gates. The automated checks run: secrets, security, dependencies, types. Red is red — it gets fixed first, not waved through.
- Decide. One senior reads the result, judges it and signs for it. Nothing goes live without that last step.
What AI does not do here
This is where we part ways with the hype. AI doesn't set architecture here, doesn't make product decisions and doesn't deploy anything on its own. Just as important: we don't trust any single check blindly — AI reviewers miss things too. That's why there are many of them, in layers, with the same human signature at the end every time. AI does the typing and the checking; the thinking and the responsibility stay human.
We're still learning too
One honest disclaimer belongs here: AI in business is younger than it looks. Today's best approach can be outdated in six months — and anyone claiming a finished playbook is selling something. That goes for us too: we keep researching where AI does and doesn't work, and we run small experiments with clear criteria — on our own work first, only then on yours. So we don't promise results. What we do promise is an honest process: start small, agree upfront what "working" means, and stop honestly when the answer is "not yet".
What you notice as a client
- Senior-level review on every change — not one audit at the end of the project.
- Security as a gate, not an agenda item. Every change is scanned before it leaves our system.
- Consistency. The same checklists, every day, deadline pressure or not.
- Speed without the usual trade. The checking adds barely any lead time, because it's automated.
And perhaps most important: if this is how we guard our own work, you know the bar we set for the automation we build for you. The same principles — start small, verifiable, a human in the loop where it counts — run through all our AI integrations.
In short
"AI-forward" isn't a sticker here but a way of working: AI builds and checks, specialist reviewers test against written criteria, and a senior signs off on every change. Curious what that discipline means for your website or web application or AI project? Book a no-strings call.