Humans and highly automated intralogistics processes

Today, a warehouse can do things that would have sounded like science fiction just a few decades ago.

Containers move autonomously through aisles. Shuttles retrieve goods from multiple levels. Conveyor systems distribute boxes at a speed that would tire a person just from watching. Robots handle palletizing, driverless transport systems move loads through warehouses, and algorithms calculate routes and priorities. A modern warehouse management system knows where each order is located, which resources are available, and which process should be initiated next.

And somewhere in the middle of it all stands a person.

Perhaps at a picking station. Perhaps in the control room. Perhaps in receiving, shipping, maintenance, or IT. Or maybe they’re no longer even in the warehouse at all, but are monitoring the system from a few kilometers away.

The interesting question, therefore, has long since ceased to be: How much work can we automate?

The more interesting question is: What work should humans still be doing in a highly automated system?

At first glance, this sounds like a difference in wording. In fact, it represents a shift in perspective. Because automation doesn’t simply change the quantity of human work. It changes its quality, its significance, and its position within the overall system. And that is precisely why humans are not becoming less important in highly automated intralogistics—they are becoming more important.

The more automated the warehouse becomes, the more human the questions become

The classic notion of automation is remarkably simple: whatever a human does, a machine can take over. Humans disappear from the process; the machine works faster, more reliably, and around the clock. This picture certainly works for individual work steps, but it does not work for an entire warehouse.

This is because a warehouse is not a machine in the strict sense. It is a socio-technical system. Goods, orders, machines, software, information, people, suppliers, customers, and time windows all influence one another. Even a technically advanced warehouse remains part of a world that is not fully automated.

A truck arrives late. An item is delivered incorrectly. A sensor provides conflicting data. An order is prioritized at the last minute. A customer changes their order. A package behaves differently than expected. A robot isn’t where it’s supposed to be. Or someone asks a very simple question: “Why has this order actually been waiting for twenty minutes?”

These aren’t failures of automation. They are the reality that automation must deal with.

The more interconnected the processes are, the less it’s about making individual work steps as efficient as possible. It’s about mastering an overall system that is constantly changing.

And this is precisely where humans come back into the picture.

The history of intralogistics can also be told as the history of the human role.

At first, people were primarily doers. They went to the goods, lifted them, transported them, sorted them, and made them ready. With each stage of automation, individual tasks were taken over by machines.

This has one obvious advantage: physically strenuous tasks can be reduced. This is particularly relevant in intralogistics. Despite digitalization and automation, employees in many places continue to perform physical labor; at the same time, new demands on perception, information processing, and decision-making are taking center stage. The Fraunhofer IML therefore investigates not only classical ergonomics but also cognitive ergonomics in intralogistics.

So humans are not disappearing from the process. Rather, they are moving up the chain within the process. They evolve from executors to supervisors, from supervisors to decision-makers, and from decision-makers to designers. At least, that is the direction in which highly automated systems can evolve.

However, this development has one prerequisite: humans must retain the ability to understand the system. After all, a decision can only be made sensibly if it is clear what information underlies it.

However, this development has one prerequisite: People must retain the ability to understand the system. After all, a decision can only be made sensibly if it is clear what information underlies it.

A dashboard with 400 metrics therefore does not in itself constitute transparency. Transparency only arises when data reveals connections. For example, when it becomes clear that an order is on hold not because order picking is too slow, but because downstream consolidation capacity has been exhausted. Or that an apparent machine downtime was actually caused by a lack of prioritization. Or that exceptionally high productivity in one area creates a bottleneck in another.

That is the difference between a system that produces data and a system that empowers people to take action.

In highly automated facilities, people are sometimes viewed as a variable factor. Machines are predictable. Software follows rules. Conveyor systems operate at defined speeds. An automated warehouse knows its capacities. Humans, on the other hand, arrive late, make different decisions, make mistakes, draw on experience, get sick, come up with spontaneous solutions, and ask questions that weren’t in any specifications.

From the perspective of a strictly deterministic system, this is inconvenient. From the perspective of a real-world system, it is indispensable.

For it is precisely where the world does not function according to plan that humans possess abilities that are difficult to formalize: an understanding of context, improvisation, experiential knowledge, intuition, and the ability to perceive contradictions.

An experienced employee can sometimes tell within a few seconds that “something isn’t right”—even before a metric turns red. He knows the sounds a machine makes. They know that a particular item is behaving differently today than it did yesterday. They recognize that a seemingly harmless deviation could become a problem later on.

This knowledge is difficult to digitize. It isn’t necessarily stored in databases. It resides in people. And that is precisely why automation should not aim to render this knowledge obsolete; it should make it usable.

The most interesting automation solutions are therefore not necessarily those where no humans are visible. They are the ones in which humans and technology play to their respective strengths. The machine takes on what it does particularly well: repeatable movements, high speeds, precise positioning, continuous operation, large volumes of data, and the processing of complex rules. Humans take on what humans do particularly well: assessing situations, handling exceptions, understanding contexts, making decisions, and taking responsibility. That sounds almost trivial, but it isn’t.

Because for this to work, the interface between humans and machines must be well-designed:

• An assistance system must not overwhelm employees with information. It must provide the right information at the right time.
• An automated workstation must not only function technically. It must also function ergonomically.
• An algorithm must not merely calculate an optimal decision. The decision must be comprehensible to humans and make sense within the process.
• And a warehouse management system must not merely display that an order is blocked. It should make it as clear as possible why it is blocked and what options for action are available.

The digitization of intralogistics is therefore always also a digitization of human decision-making.

When physical strain decreases, other types of stress can arise. A person who no longer moves several hundred boxes per hour may have to ensure that an automated system—performing thousands of movements per hour—does not fall out of sync. They must evaluate alerts, understand priorities, handle exceptions, monitor multiple sources of information simultaneously, and make decisions at the right moment—which can be highly demanding.

Research on cognitive ergonomics points precisely to this shift: technical assistance systems not only change the workflow of a task but also the demands on information processing and attention. Multitasking, time pressure, monotony, and new forms of mental strain can all be interrelated.

A highly automated work environment is therefore not automatically a people-friendly work environment. Automation is, first and foremost, a technology; whether it results in good work is a matter of design.

Perhaps the problem becomes particularly clear in an extreme scenario: Let’s imagine a fully automated facility. Everything runs perfectly. Conveyor systems, shuttle systems, robots, and software all operate in perfect sync. People are only there for the rare instances when something goes wrong.

That sounds efficient—until you ask yourself what happens to the people who work there. Anyone who spends their time solely waiting for something to break is working in a strange limbo: too important to be eliminated, but too rarely called upon to continuously apply their knowledge and skills. This is not ideal for either the person or the system. After all, a good process doesn’t just draw on human capabilities when automation fails; it integrates them from the very beginning.

People should not be the emergency exit of an automated system. They should be an integral part of its design.

This inevitably changes the role of a warehouse management system. In the past, software could be understood, in simple terms, as a digital representation of the warehouse: managing inventory, managing orders, managing storage locations. In a highly automated warehouse, that is no longer enough. Software becomes a mediator between different logics:

• The machine “thinks” in terms of movements.
• The order “thinks” in terms of priorities.
• People think in terms of situations.
• The company thinks in terms of costs, service, and delivery capability.

And the software must connect these perspectives. It must generate a comprehensible picture from a multitude of states. The goal is not to relieve people of every decision; on the contrary.

Good software should automate where rules are clear and provide support where decisions become complex.

It should not confront people with its own complexity; rather, it should make complexity manageable for them.

This is an important distinction. Because the best software isn’t necessarily the one with the most features. It’s the one that enables the right decision at the right moment.

There is yet another reason why humans remain indispensable in automated warehouses:

Automation loves the standard case. Reality loves the exception.

An item suddenly has different packaging. An order is prioritized. A gate is blocked. A shipment contains a discrepancy. A sensor reports a condition that doesn’t physically exist. A machine breaks down and an alternative route must be found.

A highly automated system must therefore be able to do one thing above all else: deal with the unpredictable.

This doesn’t mean that every exception has to be handled manually—quite the contrary. The smarter the software, the more exceptions it can automatically detect, evaluate, and, if necessary, handle on its own. But at some point, the question arises:

What should happen when no rule applies anymore?

That’s when you need someone who not only knows the rules but also understands situations—and that’s a human being. And perhaps that is precisely one of the most interesting characteristics of highly automated intralogistics: The more standard cases are automated, the more valuable the ability to handle non-standard cases becomes.

We can take this role a step further:

• A sensor measures temperature.
• A scanner reads a barcode.
• A system identifies an order.
• A person recognizes meaning.

That may sound dramatic, but in complex systems, perception is more than just data collection.

People can connect pieces of information that, at first glance, have nothing to do with each other from a technical standpoint. They pick up on signals that a system might be doing something different than its metrics suggest. The task of modern intralogistics is therefore not to keep people out of such decisions. Rather, it is to provide them with a better foundation for making those decisions.

For this to work, however, trust must also be established. Employees must understand why a system assigns a particular task. They must be able to recognize when to intervene, and they must know what happens when they correct a decision. And they must see that their experience isn’t pitted against the technology, because acceptance cannot simply be mandated through training. It arises when technology actually provides support in everyday work.

Perhaps that is the true meaning of automation. When we talk about automation, we often talk about robots, conveyor systems, sensors, and software. We talk about throughput, availability, picking rates, storage locations, and investment costs—all of which are necessary. But that describes only part of the truth. After all, a warehouse isn’t well-automated simply because as few people as possible work there. It is well automated when people and technology together form a system that functions reliably, flexibly, and in a controllable manner.

This can mean:

• That an employee lifts less.
• That they walk shorter distances.
• That they receive better information.
• That they no longer have to search for an order.
• That they can understand what’s happening more quickly when a malfunction occurs.
• That they can make decisions instead of just working through lists.

And perhaps it even means that their most important task is no longer moving things, but understanding why things move.

This also changes our view of the warehouse. It is not the sum of its machines, conveyor lines, storage locations, or software modules. It is a system in which people, technology, and processes are interconnected. Those who optimize only the technology are optimizing just one part, and those who focus solely on people are doing the same. The challenge lies somewhere in between, and that is precisely where the truly interesting questions arise:

• What should a workspace look like when the goods come to the person?
• What information does an employee need when, from their perspective, an autonomous transport system suddenly seems to be moving in the “wrong” direction?
• How much decision-making should an algorithm take over?
• When must a person be able to intervene?
• How can experiential knowledge be translated into digital processes?
• Which metrics actually show whether a system is working?
• And how do you design a warehouse so that it’s not only efficient today but can also be adapted tomorrow?

After all, this, too, is part of the reality of modern intralogistics: No system remains unchanged forever. Product ranges change, order structures change, technologies continue to evolve, business models shift, people join or leave the company, and automation is expanded. A warehouse that functions perfectly today may already be reaching its limits tomorrow.

Perhaps there is an apparent contradiction here. The more we automate, the more important it becomes not to try to automate everything. A good system needs leeway for fluctuations, for exceptions, for new requirements, for decisions—and for people. This is not an argument against automation—quite the contrary. It is an argument for a more sophisticated form of automation that does not ask, “How do we replace people?” but rather, “How can we relieve people of the tasks that machines can do better, so they can focus on what people do better?”

This requires that we not view intralogistics as a collection of isolated processes. After all, an order is not just a data record, a pallet is not just a loading unit, an employee is not just a resource, and a machine is not just a piece of technology. Everything is interconnected, and those who understand these relationships can not only make processes faster—they can make them more robust.

Perhaps that is why highly automated intralogistics will eventually be measured by a surprisingly human question:

Can people still understand what’s happening in their warehouse?

If the answer to that is “yes,” something remarkable emerges: a system in which machines handle routine tasks, software makes interrelationships visible, and people make decisions where experience, responsibility, and judgment are required. A system that doesn’t try to eliminate every uncertainty but can manage it. A system that isn’t optimized for maximum capacity at any cost, but rather for a stable flow of materials. And a system that views technology not as an end in itself, but as a tool.

Perhaps, in the end, this is also the most important perspective on automation:

Automation should not remove people from intralogistics. It should give them a different role within it.

No longer necessarily as a carrier, no longer as a messenger between two process steps. Not as someone who has to operate a machine as quickly as possible, but as an observer, decision-maker, problem-solver, and designer.

The machine can count, move, sort, and repeat; the software can plan, prioritize, and coordinate. But when it comes to understanding a complex process as a whole, one skill remains particularly valuable:

Recognizing that behind every metric lies a real operation, behind every process a connection—and behind every automated movement a system designed by people.

Other articles that might interest you:

Auf was KMU bei Automatisierungsprojekten achten sollten

TUP-Expertise zu Automatisierungssteuerung im WMS-Report von ‚Irgendwas mit Logistik‘