The Reality of Today's AI Job Market
Everyone wants a piece of the AI revolution. Companies of all sizes are scrambling to attract AI talent to avoid falling behind. This has resulted in a massive surge in job postings with titles like "AI Engineer", "Head of AI", or "Machine Learning Specialist". But there is a problem: many companies actually have no clue what they are looking for, let alone what the role entails.
HR departments and recruiters are struggling to keep up with rapid technological developments. They copy and paste requirements from other job descriptions, add the latest buzzwords, and hope for the best. As a job-seeking professional in the AI sector—whether you are just starting out or making the transition via the retraining path from developer to AI engineer—it is crucial that you can read between the lines. You need to be able to determine whether a company has a realistic vision for AI, or if they have simply published an impossible wish list.
What Job Requirements Really Mean (A Translation)
Job descriptions in the tech industry are known for being ambitious, but in the AI sector, they often reach absurd proportions. Here is a translation of common, yet problematic, requirements you will encounter in job postings.
The 'Impossible Years of Experience' Requirement
"Minimum of 5 years of experience putting Large Language Models (LLMs) like GPT-4 into production."
What this actually means: The person who wrote this job description does not have a deep understanding of the AI timeline. GPT-4 was only announced in the spring of 2023. Nobody has 5 years of experience with it yet in 2026. This often points to a lack of technical validation before the job posting went live.
Confusion Between End-User and Developer
"Expertise required in Python, PyTorch, Kubernetes, Midjourney, and ChatGPT prompting."
What this actually means: The company is lumping the building of AI systems (Python, PyTorch) and the use of commercial AI tools (Midjourney, ChatGPT) into one bucket. This is comparable to asking for a software developer who is also an expert in using Microsoft Excel. It shows that the different AI roles and their specific tasks are not yet understood internally.
The 'We Are Looking for a Wizard' Requirement
"You will be independently responsible for designing, training, implementing, maintaining, and securing our new generative AI infrastructure."
What this actually means: They have the budget for exactly one person and expect you to take on the tasks of a Data Engineer, Data Scientist, MLOps Engineer, and Security Specialist. This is a guarantee for burnout and a lack of support from management.
Three Major Red Flags in AI Job Descriptions
In addition to poorly formulated requirements, there are fundamental red flags indicating that the work environment, infrastructure, or management expectations are structurally unhealthy. Pay close attention to the following three signals.
1. The 'All-in-One' Role (The Unicorn Trap)
If the job description asks for someone who can extract and clean data, train advanced neural networks from scratch, build a front-end dashboard, and pitch the results to the board, they are looking for a unicorn. In practice, AI is a team sport. A mature AI team separates data engineering from model development and deployment. If the job description suggests that you are the entire AI department, be prepared for chaos and unrealistic deadlines.
2. The Contextless Buzzword Stack
Some job descriptions look like a bingo card. If you read sentences like: "We are implementing an AGI-inspired RAG architecture with multi-agent systems on a blockchain-backed vector database for quantum-ready inference", run away as fast as you can. This almost always means that management has read a few blog posts or listened to an expensive consultant, but has no actual business case. They are looking for technology for technology's sake, not to solve a concrete problem.
3. The Complete Lack of a Data Foundation
AI is nothing without good, clean, and accessible data. If a job description talks extensively about building smart AI solutions and predictive models, but words like 'SQL', 'Data Pipelines', 'Data Warehouse', or 'Data Governance' are nowhere to be found, the foundation is missing. You will likely find out in your first month that all the data is scattered across dozens of disorganized Excel files. You won't be building AI; you'll be acting as a glorified data cleaner.
Integrating APIs or Training Models Yourself?
A crucial distinction that is often missing in job descriptions is the difference between applying existing AI via APIs and training models yourself (Foundation Models or Fine-tuning). This difference is fundamental to your day-to-day work and has a major impact on the expected salaries within AI roles in the Netherlands.
- The API Integrator (Often AI Engineer or Software Engineer with an AI focus): The company wants to use external services like the OpenAI API, Anthropic, or Google Vertex AI and connect them to their own software or databases (for example, via RAG - Retrieval-Augmented Generation). This primarily requires strong software engineering skills.
- The Model Trainer (Often Machine Learning Engineer or Data Scientist): The company has a unique, massive dataset and wants to train or fine-tune its own model on proprietary hardware or specialized cloud environments. This requires deep mathematical knowledge, experience with PyTorch/TensorFlow, and an understanding of GPU optimization.
If the job description does not make this distinction clear, or mixes the skills of both for a project that clearly only needs an API integration, the team is likely inexperienced.
How to Assess the Company's True AI Maturity
To avoid ending up in a frustrating role, it is important to assess how mature the organization really is when it comes to data and AI. You can generally categorize a company into one of the following levels:
| Level | Characteristics | Your Role in Practice |
|---|---|---|
| Level 0: The False Start | No centralized data, no data engineers employed, management talks a lot about "integrating AI". | Building infrastructure, cleaning data, managing expectations (often dealing with disappointments). |
| Level 1: Research & PoC | Data is reasonably in order. Small experiments (Proof of Concepts) have been conducted, but nothing is in production. | Pioneering. You build the first systems, with a lot of freedom but little mentorship. |
| Level 2: Operational with Snags | Models or AI integrations are running in production. There is a budget for API costs and cloud resources. | Maintenance, optimization (e.g., improving RAG systems), scaling solutions. |
| Level 3: AI-Native (Mature) | Clear separation between Data, ML, and Software Engineering. Full CI/CD pipelines for models (MLOps). | Specialized work within a well-defined domain. Good guidance, less room for wild-west experimentation. |
For extensive knowledge on how mature companies handle the rollout of these technologies, you can learn a lot from resources on strategic implementation, such as the one found on the business AI implementation guide of our sister platform.
7 Crucial Questions for Your First Job Interview
When you have found a job opening that, despite a few minor flaws, seems interesting, it is time for the job interview. Successfully applying for an AI role is just as much about interviewing the company as it is about them interviewing you. Ask these questions to uncover the true nature of the role and the organization:
- What specific, measurable business problem is this AI role expected to solve? (If they don't have a concrete answer, you are a vanity project).
- What does the current data architecture look like, and who is responsible for the data pipelines? (Check if there is a solid foundation).
- Which AI solutions are currently actually running in production, and how many users do they have? (Separate theory from practice).
- What is the allocated budget for compute resources (cloud/GPUs) and API costs? (A lack of budget means you won't be able to build advanced solutions).
- How will the success of my projects be measured in the first 6 months? (Avoid being judged based on unrealistic expectations).
- Does this team primarily work on integrating external APIs, or do you also train internal models from scratch? (Verify if this aligns with your ambitions).
- What is the strategy regarding AI safety, privacy (GDPR), and ethics within the current team? (A mature company has a vision for this; an immature company only sees it as a problem after a fine is issued).
Conclusion: Be Critical, but Open to Growth
The AI job market is currently chaotic. Job titles rarely mean the exact same thing at two different companies, and job descriptions are often a mix of wishful thinking and buzzwords. This is no reason to give up beforehand, but it does require an extremely critical eye.
When reading AI job descriptions, you should always look for the core: Is there a data foundation? Is the problem to be solved concrete? Are the expectations humanly achievable? By recognizing the red flags and asking the right questions during interviews, you protect yourself from frustrating projects and find the companies that have made the transition from hype to actual value.