Automating internal transport: the question is not whether, but when

The technology for logistics automation is moving faster than most organisations adapt their processes. That creates an odd situation: a process written off two years ago as impossible to automate may well be within reach today, without anything about that process having changed.

So the question has long stopped being whether internal transport can ever be automated. The question is when your specific process is up, and how you notice that the moment has arrived.

The boundary shifts, your process does not

Take an internal transport process that for years carried too much product variation for an AGV. The conclusion at the time was fair: not technically feasible. Then navigation grew more accurate, sensors improved, the software learned to recognise more kinds of load, and interaction with people on the floor became safer.

The product variation is exactly the same. The process has not changed. What changed is what the technology can handle.

That principle reaches further than internal transport alone. Automated stacking and palletising, shuttle and crane systems in storage, automatic wrapping, labelling and measuring, machine vision for inspection and sorting: on each of those fronts a group of processes regularly comes within reach that was out of reach a year earlier.

The consequence is uncomfortable but simple. An automation study is never finished for good. What falls away today can be back on the table in two years.

Four forces pushing the same way

None of these four developments is new in itself. What is new is that they occur together and reinforce each other.

In the Netherlands the first point is not a feeling but a demographic fact. In its sector analysis for transport and logistics, ABN AMRO finds that over the coming years more people will leave the sector than enter it, particularly in operational roles. Forklift drivers are named explicitly, alongside order pickers, loaders and unloaders, and planners. This is not a dip that passes on its own, but a shift that keeps running.

The sector is visibly responding. In the annual study by MHI and Deloitte among five hundred supply chain professionals, attracting and retaining people tops the list of challenges, named by 58 percent. In that same study, 56 percent report that their investment in technology and automation is going up.

The sum is no longer only about saving labour

Automation was long approached as a simple calculation: what does the system cost, how many labour hours does it save, and when have you earned it back. That question stays relevant, because a solution has to be economically sound. But anyone who limits the business case to that leaves most of the value out of view.

What is an unfilled vacancy worth to you? What does absence cost, what does turnover cost, and what does it cost when you cannot expand a process because you simply cannot find the people? What is more stable output worth, or the ability to handle more volume in three years without hiring proportionally more people?

Safety belongs in that same sum, and the Dutch figures are more concrete than many people think. In the workplace accident monitor of the Netherlands Labour Authority, the forklift is involved in roughly six percent of all notifiable workplace accidents, and the pallet truck in another six percent.

There is also a connection that leads straight back to the first point. Agency workers make up four percent of everyone at work, yet are involved in eighteen percent of workplace accidents. Absorbing a labour shortage with a constantly changing pool of hired staff therefore buys you added risk. An automated step is a step where nobody has to stand under, beside or behind anything, and where induction is no longer a safety question.

What automation asks of your process

When an organisation says a process cannot be automated, there is usually a good reason behind it. A system only works when the environment around it is right, and those requirements are real.

But there is a distinction hidden in them that is easy to miss: not every condition is equally fixed.

That changes the question you should be asking. Not: does our process meet every condition for automation? But: which conditions can we solve ourselves, which do we have to accept, and which will probably be solved for us by the technology?

That is not wordplay. You do not have to wait until your whole operation is suitable. You need to know which constraint you can remove yourself and which one you have to wait out.

From one large day shift to small teams around the clock

Something else is often underestimated. Automation changes not only how the work is done, but when.

A machine has no daily rhythm, no holidays and no employment agency. As more steps run automatically, staffing around the clock becomes economically more interesting. Not because nobody is needed any more, but because a smaller group of people can support a much larger part of the day. The role shifts from carrying out to monitoring, adjusting, maintaining and improving.

That the work does not simply vanish has by now been measured more widely. For the 2026 Global AI Jobs Barometer, PwC analysed almost a billion job adverts across 27 countries and sees a labour market splitting in two. Roles in which technology takes over the routine work and human judgement moves to the centre are growing roughly twice as fast in number, with 42 percent faster salary growth, than roles that technology mainly makes easier to perform.

That study is about AI and knowledge work, not forklifts, and that difference deserves to be taken seriously. But the underlying pattern is recognisable: where technology takes over the handling, judgement becomes more valuable. Translated to the floor, whoever moves pallets today monitors an automated transport zone tomorrow. The muscle is replaced, the knowledge becomes more important.

The shift is gradual, not a switch

There is no moment at which everything flips. What you see is a shift that moves a little further year after year, with conventional and automated work coexisting for a long time yet.

That picture is confirmed from two sides. On the one hand it is clearly accelerating: in the MHI and Deloitte study, 39 percent name robotics and automation a technology with significant or greater impact, sixteen percentage points more than the year before. On the other hand, ABN AMRO is sober about what that solves. Automation, mechanisation and digitalisation can ease the shortage in part, they say, but human labour remains indispensable in daily operations for years to come.

Both pictures are true at once. A great deal is being automated, and for most organisations the end point is not an empty factory.

Automation is not a switch, it is a cycle

Because the boundary keeps moving, automation should not be a one-off project but a recurring question. In practice that comes down to seven steps you work through periodically.

The answer at step six is more often „not yet” than „no”. That is not a delay but an outcome, provided you record why it cannot be done yet and when you will look again.

Two tracks at once, and that is no contradiction

The odd thing about this subject is that it is often presented as a choice between two camps. As if you have to pick between investing conventionally or committing to automation. In practice those two tracks run alongside each other perfectly well.

Carry on buying forklifts, reach trucks and other equipment exactly as you are used to, because your operation has to run today and will for years yet. Set the study alongside it: which processes lend themselves to automation, which conditions come with them, and what is the business case. Automation is a route measured in years, so let it gradually take work away from the conventional way rather than waiting until everything can go at once.

You can buy a forklift today and at the same time investigate whether the same transport can be done by an AMR in five years. That is not a contradiction, that is looking ahead.

The risk lies in no longer looking

There is nothing wrong with concluding that a process should not be automated. Sometimes doing it by hand is simply the best solution. The problem only arises when an organisation stops investigating why.

Meanwhile technology gets cheaper, more flexible and easier to integrate, and suppliers gain experience. Anyone not following that falls behind without noticing. Not because a competitor automates their entire operation tomorrow, but because they may automate one process that you still do entirely with people. And then another.

The hardest competitor is not necessarily the existing one. A new entrant has no building already standing, no machines already installed and no processes that were once set up for a reason. They do not have to work out how to automate an existing process; they can design the process around the technology that exists today. That is a considerably stronger position, and it does not even require more technical knowledge.

The question beneath the question

Automation therefore does not start with the technology. An AGV, a robot or an automated storage system is ultimately a means. The real question starts with the operation: where does dependence on people sit, where is the physical strain, where do errors arise, which processes only run because enough people happen to be available, and which condition is holding automation back today?

Perhaps the answer is that automating now is unwise. But during that study you sometimes discover something more interesting: that the boundary has moved, and that exactly the process you said could not be done at your site two years ago is up next after all.

So the most important question is probably not whether you should automate. It is whether you know which part of your operation becomes automatable next.