Recently, we’ve been running a few searches in the Netherlands for people with experience in agentic AI. It’s not surprising that demand is increasing. More companies are moving beyond experimentation with GenAI and starting to think about how AI agents can support real business processes.
However, hiring for these roles has been more complicated than expected.
We’re seeing this from a couple of different angles. One search is with a private equity firm looking for someone to work across their portfolio, helping different businesses improve how they are using agentic AI. Another is with a large consumer goods company looking for someone to build AI agents end to end for their marketing team.
Different businesses and different use cases, but a similar hiring challenge. There are people in the market with relevant experience, but finding the right balance of technical skill, business understanding, and seniority is difficult.
The junior talent pool
At the more junior end of the market, there are a lot of technically capable candidates. Many are one or two years out of university and have moved quickly with the technology. We’re seeing people who have built RAG pipelines, worked with frameworks like LangChain or LlamaIndex, experimented with multi-agent systems, or built internal tools using LLMs.
Some of this experience is useful. A few candidates have even pushed tools into production, which is encouraging given how new this space is.
The challenge is that these roles are rarely just about building something technically interesting. They often involve working with senior stakeholders, understanding business processes, dealing with changing requirements, and explaining trade-offs to non-technical teams. That is where some of the more junior profiles are still developing.
In the right environment, they could be very effective. But they usually need structure around them. If a business hires someone junior and expects them to independently define, build, implement, and drive adoption of agentic AI across the company, that is probably asking too much.
The senior talent pool
At the senior end, candidates often come from data science, software engineering, architecture, or platform engineering backgrounds. Many have moved into GenAI and agentic AI more recently, but they understand production environments and how to build systems that work inside a business.
They are usually stronger on orchestration, monitoring, security, governance, deployment, and stakeholder management. They understand the gap between a good proof of concept and something that can be used properly by a business team.
The trade-off is cost.
In some cases, businesses end up looking at candidates with far more experience than they actually need. Deep ML expertise, heavy architecture experience, or years of low-level engineering can be valuable, but may not be necessary if the main requirement is to build internal agent workflows or marketing-facing AI tools.
That doesn’t make senior candidates the wrong option. In some situations, they are exactly what is needed. But companies need to be honest about what the person will actually do day to day.
The contractor route
One route that seems to make sense for some businesses is using contractors or consultants to get things moving, while upskilling internal teams.
A lot of agentic AI work sits in an interesting middle ground. It is technical, but it doesn’t always require a brand-new permanent hire with years of specialist experience. If a company already has capable data analysts, software engineers, or data scientists internally, they may be able to pick up areas like RAG architecture, agent orchestration, evaluation, and deployment with the right support.
This can be a pragmatic way to build capability. A contractor can help shape the first version, avoid obvious mistakes, and put some structure in place. Internal teams can then take more ownership over time.
The CV inflation problem
One thing that is very noticeable right now is how many people are trying to position themselves in this space.
Agentic AI, AI agents, autonomous workflows, multi-agent systems, and RAG are appearing on a lot of CVs. That is understandable. It is a fast-growing area and people want to show they are keeping up.
But the depth of experience varies massively.
For some people, “production” means a tool that is used regularly by a business team, monitored properly, maintained, and improved over time. For others, it means a proof of concept that worked once and never had to deal with real users, edge cases, governance, or adoption.
It doesn’t make the second type of candidate bad. Everyone has to start somewhere. But hiring managers need to understand exactly what someone has built, how it was used, who relied on it, and what their responsibility actually was.
On paper, a lot of these profiles can look similar. The difference usually becomes clear when you dig into the detail.
My thoughts
I think agentic AI hiring will stay messy for a while.
The technology is moving quickly, the terminology is inconsistent, and very few people have years of direct experience because the space hasn’t existed in this form for very long. That means companies need to be pragmatic about how they hire.
For some businesses, the right answer will be a junior builder with strong support around them. For others, it will be a senior hire who can own production and stakeholder management. For many, the best route may be to bring in external expertise and develop internal capability at the same time.
The key is being clear on what problem you are actually trying to solve.
Do you need someone to experiment and build quickly? Do you need someone to manage production systems? Do you need someone to influence senior stakeholders and drive adoption? Or are you expecting one person to do all of that?
If it is the last one, the market gets small very quickly.
Who is the best recruiter of AI staff in the Netherlands?
It’s a fair question, especially as AI roles become more specific.
For agentic AI, the difficult part usually isn’t finding people who mention agents, RAG, LangChain, LlamaIndex, or autonomous workflows on their CV. There are plenty of those.
The difficult part is knowing who has genuinely built something useful, and who has mainly experimented.
There is a big difference between a proof of concept and a system used by real business teams. Production experience means thinking about monitoring, evaluation, governance, security, edge cases, adoption, and long-term ownership. Those details matter when you are hiring for AI staff who need to make an impact inside a business.
This is where specialist recruiters can add value. Not by making the market sound easier than it is, but by understanding the difference between experimentation and delivery.
At Harnham, we spend a lot of time in the Data and AI hiring market, so we see where the real gaps are. If you’re hiring for agentic AI in the Netherlands right now, judgement matters more than volume.