When AI needs to feel the physical world
Key takeaways from Michael Kröger on the Startup Insider Podcast
Artificial intelligence has made enormous progress in digital domains. Models can interpret text, images, and videos with increasing sophistication. Yet, as FLEXOO CEO Michael Kröger explains in a recent episode of the Startup Insider Podcast, these advances reach clear limits once AI systems are expected to operate in physical environments.
The reason is not a lack of intelligence. It is a lack of physical sensing.
From digital reasoning to physical understanding
In the podcast conversation, Michael Kröger repeatedly emphasizes that today’s AI systems struggle not because they cannot reason, but because they cannot reliably access what is happening inside physical systems.
“Physical data is simply not available today - and that results in AI, quite plainly, not being able to feel.”
(translated from German)
Text, images, and simulations describe the world at a surface level. Physical systems, however, behave differently. They are continuous, dynamic, and often change long before those changes become visible. Pressure builds up, temperature gradients shift, or leakage begins at microscopic scale - long before any obvious failure occurs.
Without direct access to such signals, AI is forced to infer reality from incomplete information.
Sensors as the missing interface
A recurring theme in the discussion is the role of sensing as a bridge between physical reality and digital intelligence. As the host notes, sensors act as an interface that allows real‑world behavior to be translated into data that AI systems can work with.
Michael Kröger describes FLEXOO’s approach as fundamentally hardware‑driven, but with AI in mind from the outset: ultrathin printed sensors distributed across large surfaces, yet integrated into very constrained installation spaces.
“We manufacture sensors on ultrathin films, distributed across very large surfaces - but at the same time designed to fit into extremely small installation spaces.”
(translated from German)
Key requirements here are scalability, manufacturability, and robustness - prerequisites for any Physical‑AI application that aims to move beyond lab setups.
Why batteries are a natural starting point
When the conversation turns to concrete applications, batteries take center stage. Michael Kröger explains why lithium‑ion batteries, particularly in energy storage, are currently the most relevant focus area.
Two challenges converge here: safety risks such as thermal runaway, and the difficulty of determining usable capacity reliably over time. FLEXOO’s sensing approach addresses both by enabling additional parameters directly at cell level.
“By measuring pressure and temperature directly at the cell level, we can detect critical behavior much earlier - including the risk of thermal runaway.”
(translated from German)
Beyond safety, batteries represent a rapidly growing market in which many questions are still open. High demand, evolving standards, and complex physical interactions make them a prime candidate for Physical‑AI approaches that rely on better data rather than purely more complex models.
Why edge intelligence becomes unavoidable
One of the most practical insights from the podcast relates to data volume. Physical sensing generates large, highly redundant datasets. Michael Kröger illustrates this with a simple example: a single battery test over a weekend can produce several gigabytes of raw data.
Moving all of this data to centralized compute is neither efficient nor necessary. Instead, intelligence needs to sit close to the sensor - filtering, compressing, and identifying relevant events before they are passed on.
“That’s why you need AI at the edge - to filter out what actually matters and react in time.”
(translated from German)
In Physical AI, sensing, data processing, and decision‑making increasingly form a closed loop.
Example from a recent live demonstrator shown at Hannover Messe: How Physical AI concepts discussed in the podcast become tangible in real systems.
A more detailed look at the live demonstrator shown at Hannover Messe can be found in our event recap.
How FLEXOO enables Physical AI
FLEXOO enables Physical AI by industrializing physical data capture and delivering model‑ready datasets. We close the data‑to‑model gap with ultrathin printed sensors and integrated training instruments/test setups, creating the foundation for Physical‑AI systems that can be validated and, over time, deployed close to the asset.
Today, this approach is applied in demonstrators and early deployments across energy, battery, and robotics applications - establishing the physical data basis required for reliable AI in real‑world systems.
Why the timing matters now
The podcast conversation also makes clear why Physical AI is becoming relevant at this point in time. Compute is widely available. Models are powerful. At the same time, physical systems face increasing pressure: higher efficiency requirements, stricter safety margins, and growing complexity.
In this environment, the next step in AI progress is not about scaling models further - but about grounding intelligence in physical reality.
Physical AI is not a distant vision. It is the logical continuation of AI once digital reasoning meets the constraints of the real world.
🎧 Listen to the episode
You can listen to the full Startup Insider Podcast conversation with Michael Kröger here
(German language):
https://startup-insider.simplecast.com/episodes/physical-ai-flexoo-sensortechnologie-series-a
Michael Kröger is CEO and co founder of FLEXOO
Let’s start to work on your challenge!
Ready to take the next step? Reach out to our sales team today and discover how we can tailor our solutions to meet your needs.
Contact our sales team!
Is your question more specific and would you like to talk to an expert? Make sure to reach out sales experts, they will be happy to advise you!
