In use
AI colleagues with memory and permissions.
Role-based agents that work in the ERP/CRM, learn rules, identify themselves as AI and ask a human at clear boundaries.
Future Lab · Stage 3
Future Lab: where automation is heading and which parts of it already run today.
Not a trend report. An assessment of what will reach small and mid-sized companies by 2030, what we are already building from it, and an invitation to be the first to try it with us.
Text description of the scene: a closed cycle. At the centre, control and planning: our software or your system. It gives each stage its task, whether a person or a machine carries it out. Operated on servers in Germany, critical data runs offline. In clockwise order, six stages: (1) collection from the supplier with a route from the system and a scale on the loading crane (target state), (2) goods receipt with weighing and a photo of the material type (in use), (3) storage with portioning to the right locations (target state), (4) sales signal: sales sees the stock at once (in use), (5) order picking by a robot (target state), (6) loading and delivery with a driverless vehicle (target state). The actual data (weight, time, stock) flows back into planning and closes the cycle. A human decides on exceptions and approvals; the machines’ safety functions stay with the manufacturer. The common thread is the logic. The hardware is interchangeable.
Hands-on
The shop floor as a model: where the logic layer sits
The same thing as above, only explorable. Tilt the sheet with the pointer, click on a station, and switch between the operation as it runs today and the operation with the layer we build. The difference is not the hardware: it is there in both cases. The difference is who tells it what to do.
Move the pointer: tilt the sheet · click a station: explanation
-
01 · Warehouse
The goods are where the stock records assume they are. Which pallet is needed next is known by the order, not by the shelf.
Today: An employee walks over and checks.
-
02 · Transport robot
The manufacturer taught it to drive. From us it gets the sequence: which pallet, where to, how urgent – from the order, not from a tablet.
Today: Goes wherever a person with a tablet sends it.
-
04 · Camera
Counts, checks completeness, reads labels. The result is not an image in an archive but a number in the system and, if something deviates, a task.
Today: Records. Someone looks at it when something is missing.
-
03 · Forklift
Stays where a human is needed: heavy and irregular goods. The driver sees the same order as the robot, on the terminal.
Today: Gets shouted instructions and paper notes.
-
05 · Logic layer
The layer we build. It knows priority, exception and responsibility: what is due now, who does it, which rule applies if something goes wrong.
Today: Doesn't exist. A person makes these decisions in their head.
-
06 · Approval: human
Money, contracts, staff and every deviation out of the ordinary. Nothing happens here without approval.
Today: Decides today as well, just without the groundwork.
SubjectShop floor: handover to the logic layer
SheetZ-01
Date24 September 2026
TolerancePicture of the future, not an existing facility
Stance
Why we have an opinion about the future and write it down here
We build systems that are meant to last four or five years. Anyone who introduces an ERP today without knowing that it will have to talk to robots and cameras in 2029 builds twice. That is why we watch what is coming to SMEs and design every system so that it does not block the next step.
This is our assessment. We will be wrong in parts. But the direction is clear enough to build by it today.
By 2030
Four developments that will reach SMEs by 2030
Robots leave the halls of large corporations
Driverless transport systems, order-picking and handling robots are becoming affordable and can be rented for a monthly fee. The hardware is certified and mature. What is missing is the logic: how does the robot know which pallet customer X needs first? That is not decided by the manufacturer. It is your process knowledge. Anyone who has it written down and in the system can put the robot to work on day one. Anyone who does not puts an employee with a tablet next to it.
Cameras become inspectors, counters and witnesses
Computer vision is ready for everyday use: counting pieces, checking completeness, reading labels, reporting deviations. Not in the cloud, but on a device at the line. People only check what the camera reports as a deviation.
AI moves onto the shop floor: edge instead of cloud
Decisions are made where the data arises: at the machine, in the warehouse, in the vehicle. Without latency, without production data leaving the plant. For SMEs, this means affordable industrial AI without corporate IT.
The AI-driven company: agents work, people decide
AI agents take over whole roles: they take orders, plan routes, prepare payments, monitor assets. People move from clerk to decision-maker. This is not a vision, it runs today: AI employees work in more than 20 roles, from driver to management. By 2030 this will be normal, and the shortage of skilled workers will see to that, not enthusiasm for technology.
Today
What we are already building from it today
Being built
The cycle signal → task → execution → record.
Parts of it run today, with scales, printers and scanners as signal sources. With vibration sensors and cameras it is the same cycle; for this we are looking for pilot partners.
Predictive maintenance & industrial AIBeing built
The integration layer above the machine.
Interfaces to fleet managers (VDA 5050), assets (OPC UA) and camera systems, connected to the ERP and the warehouse. Safety functions stay with the manufacturer. We build on top.
AI for machines & roboticsIn use
AI infrastructure in Germany.
We operate memory systems, agents and business data on servers in Germany, separately for each customer, or on your infrastructure. Critical data runs offline and locally; we connect external language models only for non-critical data and only as processors under Art. 28 GDPR.
AI transparencyHonestly
We also say what we do not have yet: a completed predictive maintenance pilot in industry and robots of our own in test operation. We build the layer above the hardware, not the hardware. We are looking for partners for the first pilot projects. Become a pilot partner
For you today
What this means for you today
Three things you can do now without buying a robot:
Write down your logic.
Who decides what, by which rules, with which exceptions. This is the raw material for any automation, with or without AI. We help with this: in every project, this step comes directly after understanding the process. How we work
Bring your data into a system that has interfaces.
Excel does not talk to any robot. An open ERP does.
Start small and measure.
One AI employee for one role. One camera on one line. One asset with a sensor. Number before, number after, then onwards.
Invitation
Become a pilot partner
We are looking for pilot partners among SMEs for two projects:
Project 1
Predictive maintenance on 5–10 assets
Food business, logistics or production; existing sensors or a controller are an advantage.
Project 2
Camera-based inspection on a line or in goods issue
Counting, completeness, label checks.
What you bring: a real task, access to the data and the willingness to measure the baseline beforehand. What you get: the connection to your systems, built by us, and the result before anyone else in your industry. We discuss the terms in the first call.
Write Pilot partner
in your enquiry and name the asset or line.
Everything we build advises and reports. It does not control safety functions, does not assess people and identifies itself as AI. This also applies in the lab.
The related services: Predictive maintenance & industrial AI · AI for machines & robotics