From Automation to Architecture

The Dutch media and marketing industry has an AI problem. Not the kind that makes headlines — not deepfakes or hallucinating chatbots — but a quieter, more corrosive problem: agencies are automating content at scale without building the governance architecture to make it trustworthy.

I have spent the last year building 11 agentic workflows in Make.com — tools that screen job applicants, generate client documents, schedule multi-channel content, and deliver AI-summarised news briefings. Each one taught me the same uncomfortable lesson: the technical part is the easy part. The hard part is designing the system to be accountable when something goes wrong.

"The professionals who will be most valuable in the next five years are not those who can use AI tools. They are those who can architect the governance systems that make AI deployments safe, compliant, and commercially sustainable."

The Automation Trap

Many agencies have rushed to adopt AI primarily as a cost-reduction mechanism — using automation pipelines to generate high volumes of content with minimal human involvement. The output is fast, cheap, and increasingly untrustworthy. Audiences notice.

Research from the Reuters Institute confirms that trust in digital media continues to decline, and that readers are developing heightened scepticism toward content that feels algorithmically produced (Newman et al., 2024). This trajectory is not commercially neutral. It destroys brand equity, accumulates legal exposure, and produces exactly the kind of low-quality content that search engines are now actively penalising.

The current default is AI for speed. The standard we need is AI for governed precision. Mollick (2024) frames it directly: the professionals who create the most value in the AI transition are the ones who deploy AI inside human systems, not those who simply use AI tools.

What the EU AI Act Actually Means for Dutch Agencies

The EU AI Act (Regulation (EU) 2024/1689), in force since August 2024, binds organisations deploying AI systems that generate or personalise content at scale. Dutch businesses in the digital media space are subject to these rules whether they have read them or not.

Organisations that treat compliance as a burden face regulatory risk and reputational damage. Organisations that treat it as a competitive architecture build systems that are faster, safer, and more trusted than their competitors. The Rijksoverheid national digitalisation strategy explicitly calls for innovation that is responsible, data-governed, and human-centred (Rijksoverheid, 2024). The AI Act is the legal expression of that strategy.

Governance Layers in Practice

A Governance Layer is not a compliance checkbox. It is a set of systematic checkpoints built into the architecture of an agentic workflow that verify factual accuracy, test for bias, enforce brand alignment, and confirm legal compliance before any output reaches a human audience.

In the AI recruitment pipeline I built this year, the governance layer works like this:

The human review node is not optional. It is an architectural constraint I build in from the start, because I understand what happens when it is absent. This is Human-in-the-Loop (HITL) by design — not a feature added afterward, but a structural commitment encoded into the core of the system.

Three Principles Every AI Architect Should Build With

• 1
  Human-in-the-Loop by Design. Every agentic workflow that affects truth, representation, or user experience must have a mandatory human approval stage.
• 2
  Privacy by Design. Agentic workflows move data across nodes, tools, APIs, and external services at speed. Data containment must be designed in from the first node, not retrofitted later.
• 3
  Radical Transparency. If a piece of content or a business decision was shaped by AI, the end user deserves to know. Transparency is a structural commitment, not a communication strategy.

The 18-to-24-Month Window

The Dutch digital industry has roughly 18 to 24 months to choose between governance-by-architecture and governance-by-crisis. The professionals who lead the transition will be the practitioners who design the AI Act's obligations into the first node of their workflows, not the ones who hire consultants to retrofit compliance after a complaint, an audit, or a public failure. After that window, catching up will be measurable in lost contracts, regulatory fines and reputational damage that compounds with every AI-shaped decision a non-governed system makes.

What This Means for the Dutch Market

The Dutch digital market is well-positioned to lead on this, but only if it produces professionals who understand both the strategic context and the technical detail. From conversations with hiring managers and digital professionals across the Netherlands over the past year, one pattern emerged consistently: organisations know AI is changing how they work, but they lack the technical confidence to evaluate what "good" AI implementation actually looks like.

That gap is an opportunity. Not for AI users, but for AI architects — professionals who can walk into a Dutch agency, audit their automation setup, identify the governance risks, and redesign the system to be compliant, auditable, and commercially sustainable. The question for digital professionals building their career in this market is: which side of that gap do you want to be on?