Why the best AI managers don't write code.

The answer is a resounding no.

In fact, if you try to become a mediocre programmer now, you are already behind the curve. The code you would struggle to write today will be generated faster and more flawlessly by an AI model tomorrow. What the industry is actually crying out for is a skillset you have cultivated for years, but likely haven't applied to machines yet: leadership and technical insight.

In 2026, the engineer who wins isn't the one who builds the software, but the one who can direct it. We call this Computational Empathy. It is not a fluffy buzzword, but a hard management skill. It is the ability to translate complex, technical intent into logic that a machine cannot misinterpret.

Stop viewing AI as "software." View AI as your new intern. A very fast, incredibly well-read, but sometimes remarkably naive intern. And who is better suited to guide that intern than you?

What is "computational empathy"?

The term might sound like a contradiction. Empathy for a computer? Should we spare the feelings of a server rack?

Of course not. Computational Empathy is not about feeling with the robot. It is about having empathy for the limitations of the robot. It is the realization that a Large Language Model (LLM) or an autonomous agent has no 'common sense,' possesses no intuition, and does not inherently 'sense' the context of your construction project in the Botlek area.

Think back to the first time you had to mentor a fresh graduate. If you told that junior: "Take a look at that bridge construction," you probably got a blank stare in return. What did they need to do? Inspect the welds? Calculate the foundation? Assess the aesthetics?

As a senior, you learned quickly that you had to be specific: "Check the moment capacity of the girders on axis 3 according to the Eurocode."

That is Computational Empathy. It is the art of extreme clarity. Many engineers think "Prompt Engineering" involves learning magic spells. That is nonsense. Prompting is nothing more than the highly accurate delegation of tasks. If you get frustrated because an AI tool doesn't understand you, it is rarely the tool's fault. It is usually a lack of context in your instruction.

The 3 pillars of agent management (without code)

To transform from a traditional engineer into an AI manager, you don't need to go back to school for an IT course. You need to project your existing project management skills onto your digital workforce. This rests on three pillars.

Pillar 1: Context loading (the onboarding)

A new AI session is like a new employee walking in on their first day. They know nothing. They don't know the client, they don't know the project history, and they don't know which standards you must comply with.

Many engineers make the mistake of asking a question immediately: "Write a report on material fatigue." The result is generic, superficial, and often unusable.

You must onboard the agent. In jargon, this is called persona setting. Before you assign the task, you provide the context.

• The Bad Approach: "Calculate the wind load."
• The Computational Empathy Approach: "You are acting as a Senior Structural Engineer, specialized in high-rise buildings in the Netherlands. We are working on a renovation project in Groningen, so earthquake resistance is a factor. You must strictly adhere to the NPR 9998 standard. The target audience for your output is the project developer, so keep the language technically correct but accessible. Do you understand this role?"
• Only when the agent confirms do you give the actual assignment. This costs you two extra minutes of typing, but it saves you three hours of rewriting.

Pillar 2: Logical decomposition (the project plan)

This is where your experience as an engineer is worth its weight in gold. AI models have a tendency to 'hallucinate' (make up facts) if a task is too big or undefined. If you ask an AI to "write a full specification for a water treatment plant" in one go, you will get nonsense.

The solution is logical decomposition. In construction and engineering, we call this a Work Breakdown Structure (WBS). You chop the big problem into small, manageable chunks.

Instead of one giant prompt ("Design an HVAC system for this office"), you conduct a dialogue in steps:

1. Step 1: "Based on this floor plan (upload), give me an estimate of the required cooling capacity per room based on volume and glazing."
2. Step 2: "Given this capacity, propose three possible system configurations that meet BENG (Almost Energy Neutral Building) requirements."
3. Step 3: "Select option 2. Now generate a list of specific components required for this configuration."

This is not programming. This is process control. It requires you, the expert, to understand the order of operations. A junior doesn't know which steps are needed; you do. That is why you are the driver, and the AI is the engine.

Pillar 3: Verification (the code review)

The old proverb "Trust is good, control is better" is the core of working with AI.

An AI agent is confident, even when it lies. It will present a completely fabricated formula with the same conviction as a correct one. This is where the professional separates themselves from the amateur.

The amateur copy-pastes the output directly into a report. The professional applies verification. Treat the AI's output as the work of an intern. You don't blindly assume the calculation is correct. You grab your calculator, use your common sense, and check the assumptions.

• Is the output logical?
• Are the units correct (did it mix Imperial and Metric)?
• Does the AI refer to a NEN standard that actually exists?

Your value lies no longer in making the first draft, but in validating it. You shift from creator to auditor. This is a role that fits seniority much better. It removes the mind-numbing typing work and leaves the high-value thinking work.

Why this is critical for the Dutch sector

Why do we emphasize this at Randstad? Because the Dutch engineering sector has unique characteristics that align perfectly with this new way of working.

1. The Culture of Efficiency and Directness The Dutch are known worldwide for their directness. Niet lullen, maar poetsen (Don't bullshit, just work). In human interaction, this is sometimes perceived as blunt, but in interaction with AI, it is a superpower. AI models function best with unambiguous, direct instructions without 'social noise.' The Dutch communication style is, ironically, perfect for digital transformation.
2. The Shortage of Technical Talent We cannot ignore it: there is a screaming shortage of engineers in the Netherlands. Vacancies remain open for months. The workload on existing teams is increasing. We cannot simply conjure up thousands of new senior engineers. What we can do is equip current engineers with 'exoskeletons' for their brains.
   
   An engineer who masters Computational Empathy is a so-called Force Multiplier. He or she does not do the work of one person but facilitates the work of three. By outsourcing repetitive tasks (writing reports, analyzing data, drafting emails, checking code) to agents, you free up time for what you were actually hired for: complex problem solving and innovation.

Conclusion: your future role

The fear that AI will take over your job is unfounded, as long as you are willing to let your role evolve. The engineer who refuses to work with AI will be replaced—not by the AI itself, but by the engineer who effectively utilizes it.

You don't need to book a Python course tonight. You don't need to enroll in a Data Science degree. What you do need to do is look critically at how you communicate. Are you able to structure your immense technical knowledge in such a way that a non-human entity can execute it?

That is the core of Computational Empathy. It is the bridge between your human expertise and the computing power of the machine.

At Randstad, we see a shift in the demand from our clients. They are no longer just looking for the best calculators, but for the best 'connectors.' Professionals who understand how to manage a digital team. Are you ready to take that step?