Today
Condition data piles up in controllers and monitoring tools that hardly anyone opens. The repair history sits in the maintenance system, spare parts in the ERP, inspection rounds in spreadsheets. Because nobody brings it together, failures still come as a surprise, usually mid-shift, and the part is not in stock.
An example
At Rabenhammer Präzisionsteile, a cylindrical grinder running three shifts grinds the bearing seats of the W Ø 40 h6 shafts. For three weeks the vibration at the grinding spindle has been rising, and the bearing temperature is four degrees above normal. The agent finds two earlier spindle failures with the same pattern, which stopped the machine five and seven weeks later, and sees that no exchange spindle is in stock. It proposes ordering one now and swapping it during the planned stop in two weeks. The maintenance manager approves, and purchasing places the order the same day.
Your data, wherever it lives
Vibration, temperature and current draw sit today in controllers and monitoring tools, inspection rounds in spreadsheets, breakdowns and repairs in maintenance tickets, spare parts in the ERP. demi reads, understands and links them for every machine, with your rights. The AI sees in minutes which machine behaves unusually and what followed such signs before, and your maintenance team decides on the intervention.
Where the data lives today
- Machines & sensors
- Tickets
- ERP
- Office files
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: machines and sensors, the maintenance system, the ERP and spreadsheets. demi replaces nothing and joins measurement series, fault messages, repair orders, replaced parts and stock for every machine. Orders and reservations are written back only after approval.
- 2 · PermitPermission Layer
Condition data and history are visible to whoever looks after the machine: the plant's maintenance team and the line's shift leaders. An external service technician sees only the machine you release for their job. An authorized person approves interventions and part reservations, and that is recorded.
- 3 · ActIntelligence Layer
The agent learns the normal state of every machine from vibration, temperature, current draw and pressures and compares deviations with the patterns that preceded earlier failures. It backs every risk with the readings and the most similar cases from the history. It proposes the date and the part; maintenance makes the decision.
How it runs
- 1
demi links measurement series with breakdowns, repairs and replaced parts for every machine.
- 2
The agent learns the normal state and the patterns that preceded earlier failures.
- 3
Every week it rates the failure risk of each machine and backs it with the most similar cases.
- 4
For machines that stand out, it proposes the intervention in the next planned stop, with the spare part from stock or ordered in time.
- 5
Maintenance checks and approves; the order and reservation go into the system.
What changes
Fewer unplanned stops
Interventions in planned downtime, not on the night shift
Spare parts ready before they are needed
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+120 hfreed per year
Assumptions and calculation
demi spots unusual machines earlier and plans the work and spare part; you decide on the date.
5 h → 4 h per case
How we calculate
10 breakdowns × 5 h × 20% × 1.0 (manual) × 12 months = 120 h
Value per year, before demi costs
Freed per year+120 h
Assumption: labor cost of $48.62 per hour in US manufacturing in June 2026 (BLS), plus workplace, IT and overhead.
= Value per year, before demi costs
+$7,800
gained per year
120 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
Do we need new sensors for this?
Usually not. demi uses what is already recorded: controller data, condition monitoring, inspection rounds in spreadsheets, energy meters. If an important signal is missing, the agent shows which extra measuring point would help most. demi does not access the controls.
How reliable is a prediction?
It is a reasoned assessment, not a certainty. For every risk the agent shows the readings and the earlier cases it rests on, and how often this pattern actually led to a failure. Your maintenance team decides what to do.
What if our history holds hardly any failures?
Then the agent starts with the normal state and reports when a machine behaves differently from usual, even without an earlier failure as a model. Every confirmed breakdown and finding sharpens the patterns. It also reads the free text in old repair orders.
Does this replace our maintenance plan?
No, it complements it. Fixed intervals stay where regulations or the manufacturer require them. Where the condition shows that a part is due earlier or lasts longer, the agent proposes an adjustment, backed by readings and earlier cases.
