Today
Customers send in samples of their thermal oil, cooling water or cleaning bath; the lab measures, and an application engineer writes every recommendation by hand. To do so he hunts for limits in the data sheet, earlier samples in folders and e-mails and the customer's plant in the CRM. The customer waits a week, and nobody sees how a plant's values develop over the years.
An example
Orvelon supplies the thermal oil for the press of a composite panel maker, 4,500 liters at 290 °C flow temperature. The annual sample shows a flash point of 176 °C instead of last year's 214 °C and slightly higher carbon residue. The agent lays the six earlier annual samples alongside: up to last year the flash point held steady between 214 and 220 °C, and in an e-mail from April the customer mentions replacing the circulation pump. It proposes local overheating from too little flow through the heater as the cause and recommends measuring the flow, removing the low boilers via the expansion tank and testing again in three months instead of changing all the oil at once. The application engineer adds a note on the pump curve and releases the recommendation the same afternoon.
Your data, wherever it lives
What a recommendation needs is scattered today: readings in the lab system and in test reports as PDFs, earlier samples in folders and spreadsheets, plants and contacts in the CRM, deliveries in the ERP, the customer's notes in e-mails. demi reads, understands and links all of it per customer and plant, with your rights. AI rates every sample on the day it is measured and drafts the recommendation; your application engineer releases what the customer receives.
Where the data lives today
- Databases
- PDFs
- Folders & drives
- Office files
- E-mails
- CRM
- ERP
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: lab system, ERP, CRM, e-mail and test reports as PDFs. demi replaces nothing and links every sample to customer, plant and product, to the limits and guide values from data sheet and standard, to all earlier samples from the same plant and to the deliveries in the ERP, so you see when and how much was topped up.
- 2 · PermitPermission Layer
Each customer sees only his own samples and recommendations, field staff and application engineers only their customers. Formulations and internal warning thresholds stay in the lab. A recommendation reaches the customer only after the responsible application engineer has released it.
- 3 · ActIntelligence Layer
The agent compares every lab value with its limit and guide value and with the plant's history and recognizes patterns, such as falling flash point with rising carbon residue or falling inhibitor with rising iron. From this it drafts the recommendation: top-up with quantity, partial or full change, plant check, next sampling date. Quantities are computed by fixed formulas from fill volume and target concentration; the language model only explains.
How it runs
- 1
The sample arrives in the lab with customer, plant and sampling date; the readings go into the lab system.
- 2
demi links every value to its limit and guide value and to this plant's history.
- 3
The agent rates the sample and drafts the recommendation with quantities, reasons and trend charts.
- 4
The application engineer checks, changes what is needed and releases it.
- 5
The customer gets report and recommendation the same day; a top-up goes into the ERP as a quote or order.
What changes
Recommendations on the day of measurement, not a week later
Trends over years visible before a plant is damaged
Top-ups and changes become plannable orders
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
+288 hfreed per year
Assumptions and calculation
demi rates the lab values and drafts the recommendation; the application engineer checks and releases it.
30 min → 18 min per case
How we calculate
120 customer samples × 30 min × 40% × 1.0 (manual) × 12 months = 17,280 min = 288 h
Value per year, before demi costs
Freed per year+288 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
+$18,720
gained per year
288 h ÷ 1,760 h per role = 0.2 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
Which samples can demi evaluate?
Any for which your lab supplies values and for which limits or guide values exist: thermal and insulating oils, cooling and boiler water, cleaning and pickling baths, coolants or disinfectant solutions. The agent works with the parameters your lab measures, such as flash point, acid number, viscosity, conductivity or inhibitor content.
Where do the limits for customer samples come from?
From your data sheets and test specifications, from standards such as DIN 51529 for used heat transfer fluids and from the plant manufacturer's rules. Every limit in the report names its source. Your lab sets internal warning thresholds; demi only applies them.
How does the agent compute a top-up?
By fixed formulas from the plant's fill volume, the measured concentration and the target value, as your data sheet specifies. The fill volume comes from the plant record or the first fill in the ERP. If it is missing, the agent asks instead of estimating.
Can the customer see the history of his plant?
Yes, if you release it: with every report he gets the trend charts of his plant across all samples, as a PDF or in your customer portal. He then understands why you recommend a change and plans it into his next shutdown in time.
