Kronehoefstraat 85
5612 HL Eindhoven
BTW NL863902509B01
The future of work in an AI-driven world
qw
vc
mshm
.svg)
AI mostly replaces individual tasks in SMEs, not entire jobs: roles shift from doing to reviewing. When you help your team combine AI with their existing expertise, they get more done without extra hires, which is exactly where the gain lies in a tight labor market.
Many SME directors avoid the AI conversation with their team. Not because they think the topic is unimportant, but because the question everyone is asking is the wrong one. “Will AI cost jobs?” only creates unease, and gives you no answer you can actually use in your own business.
Investor Marc Andreessen put it sharply in January 2026, in conversation with Lenny's Podcast: “everyone talks about job loss, but you need to look at task loss: the job outlives the individual tasks.” That distinction is exactly where most conversations about AI and staff go wrong.
A job is made up of dozens of tasks. Some of them disappear or change significantly. Others stay unchanged, human work. Start the conversation at “which jobs disappear” and you get panic. Start it at “which tasks change” and you get a concrete agenda.
For a Dutch employer in administration, construction, real estate or staffing, that difference is not academic. It determines whether your team sees AI as a threat or as something that makes their workday lighter. And that, in turn, determines whether they work with you or against you.
This article is therefore not about whether AI costs jobs. It is about which tasks in your business are shifting, what stays with your people, and how you have that conversation without causing panic.
Andreessen makes a second observation in the same conversation that matters just as much: someone who combines two skills is worth more than twice as much, and someone who combines three is worth more than three times as much. AI speeds up that combination for everyone, not just specialists.
Concretely: the tasks that shift first are the ones without judgment calls. Retyping, looking up data, summarizing, drafting standard correspondence, building a first version of a report. These are exactly the tasks an AI agent handles well: software that carries out a series of steps on its own based on an instruction, instead of answering a single question.
Under the hood, that usually runs on a large language model (LLM), a language model that reads, understands and produces text. For a planner at a staffing agency, that means: the model reads a CV, summarizes it, and drops it into the right template. The planner judges whether the match is right.
The World Economic Forum predicts that automation will displace 92 million jobs worldwide by 2030, against 170 million new roles (World Economic Forum, January 2025). That is not a net loss of jobs: it is a shift in task content, with more work overall, not less.
For an SME without its own IT team, that global figure matters less than the question of which of those tasks sit in your own process. Walk through an average employee's workday and count how many minutes go to retyping and searching. That is the starting point, not the global number itself.
Not every task shifts. An AI agent delivers a draft answer, not a decision someone is accountable for. Three categories remain structurally human work.
Judgment on exceptions. A quote that falls just outside the standard process, a customer who calls angry about an invoice, a building permit with an unusual detail. AI recognizes the pattern, a human decides what happens.
The client relationship itself. An accountant preparing a business owner for a difficult conversation with the bank does more than check numbers. You don't build that trust with a language model.
Context that is nowhere on record. The real-estate agent who knows this seller actually wants a quick sale, even though it isn't written down anywhere. That knowledge lives in people's heads, not in systems, and will for the foreseeable future.
A fourth category is less obvious: taking responsibility when something goes wrong. An AI agent can flag an error in a file, but the decision to call a customer about it and make it right stays with a person who is accountable for it.
This is exactly why “AI replaces people” is the wrong summary. AI takes over the repeatable part. The non-repeatable part, where most of the tension and value in a client relationship sits, stays with your team.
It only gets concrete when you look at individual roles. The examples below are illustrative: a common pattern from conversations with SMEs, not an existing client case. They show how a role shifts, not disappears.
In all three cases, the role doesn't disappear, only its low-value, repeatable part does. What remains actually demands more judgment, not less.
If tasks disappear, why do people become more valuable instead of cheaper? The answer lies in timing and skill combination.
OECD research shows that employees with AI skills earn 21 percent more on average than colleagues without those skills, precisely because AI knowledge is strongly complementary to existing expertise (OECD, April 2025). Not instead of what someone already knows, but on top of it.
Andreessen calls the timing of AI “remarkably good”: after fifty years of slow productivity growth and declining population growth, AI arrives right when the remaining workers are becoming scarcer, not more plentiful. That makes them more valuable, not more replaceable.
Translate that to your own business: the colleague who now manually stitches two systems together becomes more valuable the moment AI supports that work instead of doing it fully by hand. Not because she has less work, but because she gets more done in the same time.
For the Dutch labor market, this is not theory. UWV reports that employers still face a persistently tight labor market (UWV, 2025). Growing by hiring simply isn't working in many sectors. Growing by getting more out of your existing team, with AI as a lever, is. We call this connecting your systems and knowledge into one AI business brain: your team works from the same information, without knowledge being stuck in one person's head.
The conversation with your team doesn't need to be a big announcement. It works better as a series of short, concrete conversations.
Start with the task everyone hates. Simply ask: which recurring work costs the most time and delivers the least? That answer comes faster than you'd think, and it's almost never the task management assumes.
Let people point out what can go. Whoever makes the proposal doesn't feel like the victim of a decision from above. That difference decides whether AI lands as a tool or as a threat.
Be explicit about what you won't do. Say out loud that this is not about layoffs, but about less retyping and more time for work that actually needs judgment. Vague reassurance doesn't work, a concrete boundary does.
Have this conversation per team, not as one plenary session for the whole organization. A planner has different concerns than an account manager, and a tailored conversation feels less like an announcement and more like thinking it through together.
Plan a follow-up after the first few weeks. Ask concretely what's going better and what still doesn't sit right. That repetition shows it isn't a one-off announcement, but a way of working that stays.
Without your own IT or AI team, you don't start with a platform, but with three processes you can point to yourself.
Choose processes with a lot of repetition, a clear input and output, and a measurable number of hours per week. Checking invoices, drafting quotes, preselecting CVs: these are typical candidates. In the first round, avoid processes that revolve around judgment or client contact.
In the Netherlands, a minority of businesses still use AI technology structurally: among micro-businesses with two to nine employees, that was 13.8 percent in 2025, against 29.8 percent for businesses with 10 to 249 employees (CBS, 2026). Whoever takes a first step now is still ahead of most of the sector.
Measure in hours, not in technology. Not “have we implemented AI” but “how many hours a week is this task shorter now.” That is the number that convinces your team, and one you can repeat on the next process next quarter.
Share that number back with the team that pointed out the task. Seeing that their input genuinely saves hours is what makes the next conversation easier than the first.
This is exactly where an outside perspective helps. In our AI Strategy & Roadmap, we first map out which tasks in your organization are really shifting, before we build anything. For running those first processes, we look at intelligent process automation: workflow automation that takes over the repeatable part of a task, so your team is left with the part that needs judgment. That way your business grows with your current people as digital workers alongside them, instead of with a bigger payroll.
Usually not directly. AI mostly takes over repeatable tasks within an existing role, not entire functions. World Economic Forum research expects more new roles globally than disappearing ones (World Economic Forum, 2025), but that does require employers to actively bring their team along in the shift.
Retyping, looking up data, summarizing and drafting standard correspondence are usually first in line. These are tasks without judgment calls, with a clear input and output, exactly where an AI agent excels.
No, that's usually not necessary and not wise either. Most Dutch SMEs are actually dealing with a tight labor market (UWV, 2025). AI lets your existing team get more done, which makes growth possible without extra hires.
Start small and per team, not with one big announcement for the whole organization. Ask which task costs the most time and delivers the least, and be explicit that the goal is less retyping, not fewer people.
Roles with a lot of repeatable, administrative work change fastest: administration, planning and first-line client contact. The core of the role, such as client relationships and judgment, usually changes less than the execution work around it.
Judgment on exceptions, the client relationship itself, and context that is nowhere on record. Those are three categories that remain human work for now, and that become more important as the routine work around them disappears.
Choose three processes with a lot of repetition and a measurable number of hours per week, and start small. At The Agentic Group, we often start with an AI strategy that pinpoints exactly which tasks shift first, before anything gets built.
With a well-chosen, repeatable process, the first time savings are often noticeable within a few weeks. Measure in hours per week per task, not in technology, and you'll quickly see whether a process is worth the investment.
We map out which work in your business can be automated and what that delivers for your team. No reorganization, just a concrete picture of what's shifting.
.svg){kind=icon}