Today
A customer reorders a shade last made ten years ago, or sends a standard that resembles an old colour. The recipe from back then sits on an index card, in a recipe book or in a spreadsheet, the measurements in the colour measurement software, the customer's approval in an e-mail. So the lab starts its lab dips almost from scratch, and every round costs a day.
An example
At Fibrana, a customer reorders a coated PA 6.6 fabric of 220 g/m² in an olive shade last made in 2014. The recipe was on an index card, and the swatch on the sample card has faded. Among 6,300 recipes read in, the agent finds three olive shades on PA 6.6 with measurements; the closest is at ΔE 0.6 to the customer's approval back then. One of its three dyes is no longer available, so the agent names the replacement from a 2021 recipe approved on the same fabric. The lab manager dyes a lab dip from it, and the customer approves the first sample.
Your data, wherever it lives
Colour recipes sit today on index cards and in old recipe books, in spreadsheets and in the colour kitchen's database, measurements in the files of the colour measurement device, customer approvals in e-mails, articles and orders in the ERP. demi reads, understands and links them per recipe, with your rights. For a colour request the AI finds the closest proven recipe in minutes, and your lab decides what to dye with.
Where the data lives today
- Scans & paper
- Office files
- Databases
- Machines & sensors
- E-mails
- ERP
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
The case needs data from the recipe archive, the colour kitchen, colour measurement, the ERP and the mailbox. demi replaces nothing and turns every recipe into a record: material or substrate, dyes and pigments with their shares, auxiliaries and the process with temperature, time and pH. Each recipe is linked with its measurements (reflectance curve and L*a*b* values), the article, the customer, the approval and the orders it ran in.
- 2 · PermitPermission Layer
Recipes are your know-how: only the lab, the dyehouse and development see them, dye prices only purchasing. Sales sees what a customer approved, without the recipe. A recipe goes into production only after the lab manager releases it; recipes stay in the house, and the customer receives the sample with its measurements.
- 3 · ActIntelligence Layer
The agent also reads handwritten index cards and old recipe books and maps old dye names to today's products. For a new request, whether a physical standard, a measurement or an old article, it finds the closest recipes on the same material. demi calculates colour differences with the formula you use, such as CIEDE2000, also under a second illuminant to detect metamerism. It shows which dye is no longer available or is on your list of restricted substances and names the replacement from a proven recipe. The language model only explains.
How it runs
- 1
Index cards, recipe books, spreadsheets and colour measurement exports are read in, even whole archive boxes.
- 2
The agent records each recipe with material, dyes, process and measurements and links it to article, customer and approval.
- 3
When a request arrives, it finds the closest recipes on the same material and calculates the colour differences.
- 4
It flags dyes that are no longer available or restricted and proposes the replacement from a proven recipe.
- 5
The lab manager chooses the starting point, dyes the lab dip and releases the recipe for production.
What changes
Repeat shades from a proven recipe instead of from zero
Fewer lab dip rounds until customer approval
Decades of colour knowledge findable, even when the colourist retires
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
+216 hfreed per year
Assumptions and calculation
demi finds the closest proven recipe with measurements for each colour request; you dye and release.
90 min → 63 min per case
How we calculate
40 colour requests × 90 min × 30% × 1.0 (manual) × 12 months = 12,960 min = 216 h
Value per year, before demi costs
Freed per year+216 h
Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.
= Value per year, before demi costs
+€14,040
gained per year
216 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
Can demi read old index cards and recipe books?
Yes, including handwritten entries, old dye names and dyehouse abbreviations. Every value read shows the image snippet it comes from. Where a quantity or a name is uncertain, the agent asks for confirmation.
Where do the colour measurements come from?
From the exports of your spectrophotometer and colour software, from lab reports or from spreadsheets. If an old recipe has no measurement, the agent shows that; if the sample still exists, the lab measures it again. demi does not read colour values from photos of sample cards, because camera and light distort them.
Does this replace our colour matching software?
No. Colour matching software calculates a recipe from measurements; demi gives it the best starting point from your own archive, together with approval, order and production experience. Corrections stay where you make them today.
Does it also work for stains, paints and masterbatches?
Yes. The principle is the same: recipe, material, process and measurements are linked, and the search finds the closest proven match on the same substrate: for oak, recipes proven on oak count, not those on beech; for PP, not those on PE. Your lab decides which features drive the comparison.
