Digital Transformation

Are 'dark factories' real? A podcast series on the future of the factory floor

Are dark factories real? Yes, but fully lights-out plants are rare. For most manufacturers, the mature path is a grey factory: partial automation where robots and people share the work. Shop Floor Stories: Dark Factories is a three-part podcast that separates hype from reality with scientists, a founder scaling automation, and a labor leader.

Gepubliceerd op:
01 July 2026
Bijgewerkt op:
01 July 2026
Delen:

A dark factory (also called lights-out manufacturing) promises a plant that runs with no people on the floor: robots working through the night, lights off, production uninterrupted. It is one of the most talked-about and most misunderstood ideas in manufacturing.

Are dark factories real? Yes, but only for a narrow slice of high-volume, low-variation production. For most plants, the realistic path is not full autonomy but a grey factory: partial automation where people and robots share the work. That is the question we put to three very different perspectives in Shop Floor Stories: Dark Factories, a new three-part podcast series:

  1. The scientists who actually build the robots,
  2. A founder scaling automation inside real factories,
  3. A labor leader who represents the people on the floor.

The goal is not to sell a lights-out utopia but to separate what is real from what is hype, and to understand what it genuinely takes to automate well.

Episode 1: the science, with UGent professors Francis Wyffels and Andreas Verleysen

To kick off the series, UGent Robotics and AI professors Francis Wyffels and Andreas Verleysen reframe the term "Dark Factory" itself. "Dark doesn't need to be dark," Wyffels explains. "It can also be about improving existing lines, while people can still be around working together with these robots." Their research focus has shifted from high-volume, low-variety production toward the high-mix, low-volume reality where the robot has to adapt to the task rather than the other way around.

"We are highly overestimating what humanoid robots are currently capable of doing."
Andreas Verlaenen

On the humanoid hype, both professors are blunt. Verlaenen notes that the viral demos often mislead: "If you look closer at these commercials, you can see that this robot is often being operated from a distance by a human." Language models had the entire internet to learn from. Robots do not. Which is why humans stay in the loop. Verlaenen adds: "Our hands are magical. So we will need this human in the loop to collaborate with."

"If we take the current tech and just keep on doing what we're doing now, then I would say we don't have general purpose robots in the first 100 years."
Andreas Verlaenen

The takeaway for manufacturers is practical, not pessimistic. Full autonomy is rare and far off. Partial automation, with people and robots sharing the work, is the realistic and mature path.

Asked whether robots still need instructions, Wyffels didn't hesitate: "They need working instructions. They also need a way to understand these working instructions. And then they also need training data to learn how to do the task." Instructions don't disappear when robots arrive. They become the shared foundation that both people and machines run on, and someone still has to capture how the work is done and keep it standardized. That is exactly the layer Azumuta builds.

Clip from Shop Floor Stories episode 1 with Francis Wyffels and Andreas Verleysen, UGent Robotics and AI.

Listen to the full episode here →

Episode 2 (coming soon): scaling automation, with Jonathan Berte (Robovision)

Next, we move from the lab to the factory. Jonathan Berte, founder of Robovision, has spent years putting vision AI to work on real production lines.

Episode 2 digs into why high-mix, low-volume "dark" factories are now technically possible, why most factories still stay "grey" rather than going fully dark, and why partial automation is often the smart engineering choice, not a compromise. Where Episode 1 covers the science, Berte brings the hands-on reality: what scaling automation looks like when the products keep varying.

Episode 3 (coming soon): the human side, with Stijn Gryp (ACV)

Finally, the people. Stijn Gryp, national secretary at the ACV union and a social scientist, brings the perspective the tech conversation often skips: what automation does to jobs, skills, and "workable work."

Episode 3 covers upskilling, social dialogue, and the robot-tax debate: the questions that decide whether automation enables a workforce or leaves it behind.

Listen now, and follow for the next two episodes

Episode 1 with Francis Wyffels and Andreas Verleysen is live now on YouTube, Spotify, and Apple Podcasts. Follow the show to catch Episode 2 with Jonathan Berte and Episode 3 with Stijn Gryp as they drop in the coming weeks.

Frequently asked questions

Quick FAQs to get you up to speed

A dark factory, also called a lights-out factory, is a plant that runs production with no humans on the floor. Because no one needs to see, it can in principle operate in the dark. It suits narrow, repetitive, high-volume tasks with stable inputs.

Yes, but they are rare. As UGent robotics professor Francis Wyffels notes, "dark doesn't need to be dark," and people often still work alongside the robots. Fully autonomous, lights-out plants exist only for a narrow slice of high-volume, low-variation manufacturing.

A dark factory runs with no people present. A "grey" factory uses partial automation, where robots and people work together. Robots handle predictable, high-precision tasks, and humans handle judgment, flexibility, and exceptions. Grey is the mature choice for most plants.

A grey factory uses partial automation: robots and people work side by side on the same production lines. Robots take repetitive, high-precision tasks; humans handle exceptions, judgment, and dexterous work. Most manufacturers today operate in grey mode rather than fully lights-out.

Not soon. UGent researchers estimate a true general-purpose robot is at least a decade away, and at current data-collection rates, possibly far longer. For now, robots take over repetitive tasks while humans stay in the loop for dexterous and judgment-based work.

Digitize and standardize human work first, and start capturing operational data now. Robots scale a process; they do not fix one. Make work consistent and visible before automating, then keep people in control of the rest.

Sluit u aan bij de digitale werkvloerrevolutie!