Skip to content
Planning & productionPlastics & PackagingChemicals & Process IndustryFood & BeverageMedical Technology+2

Compare every batch with the best one

demi lays every batch over your best ones and shows which deviation in temperature, dosing or time costs yield and quality.

Today

Same recipe, same reactor, and still one batch runs smoothly while the next takes hours longer, needs correcting or fails the lab test. The difference hides in thousands of curve points in the process control system that nobody lays over each other batch by batch. Lab results, raw material lots and shift notes sit in other systems, so every explanation stays a guess.

An example

At Orvelon, a batch of the AR 240 cleaner in the 6,300 l reactor takes between 5 and 9 hours, and one batch in six needs its alkalinity or clarity corrected. The agent lays 160 batches from the past year over the twelve best: in the slow ones the caustic soda went in over 25 instead of 60 minutes, the temperature rose above 70 °C, and the surfactant was added before the mix had cooled below 45 °C. On top of that come three lots of complexing agent with less active substance. Process engineering sets a dosing ramp and a maximum temperature for adding the surfactant, and quality approves the revised manufacturing instruction.
Fictional company, typical workflow.

Your data, wherever it lives

Your batch curves sit today in the process control archive, phases and batch records in the MES, lab results in the lab database, raw material lots and yield in the ERP, shift notes in spreadsheets. demi reads, understands and links them per batch on one timeline, with your rights. The AI lays every batch over your best ones in minutes and shows what sets them apart, and your process engineers decide what becomes the rule.

Where the data lives today

  • Machines & sensors
  • MES
  • Databases
  • ERP
  • Office files
Data LayerOne view of everything, with your rights

How demi solves it

Three layers in this order: connect the data first, then the rights, then the AI.

  1. 1 · ConnectData Layer

    Data sources in this case: process data from the process control system, batch records and phases from the MES, lab results from the LIMS, raw material lots and yield from the ERP. demi replaces nothing and puts temperature, pressure, dosing quantities, agitator power and phase times of every batch on one timeline, linked to lab result, raw material lots and shift. Every batch can then be compared phase by phase with the best ones.

  2. 2 · PermitPermission Layer

    Production, process engineering and quality see curves and deviations; recipes and reference curves are visible only to the roles you approve. A new rule for dosing, temperature or time applies only once process engineering and quality release it; every release is recorded.

  3. 3 · ActIntelligence Layer

    The agent derives the reference curve per phase from your best batches, lays every new batch over it and marks where and how far it deviates. Across all batches it calculates, with traceable statistics from the recorded data, which factors go along with yield, lab results and batch time. The language model only explains the result.

How it runs

  1. 1

    After each batch, demi reads the process curves, batch record, lab results and raw material lots.

  2. 2

    The agent lays the batch over the reference curve from your best batches, phase by phase.

  3. 3

    It marks every deviation in temperature, dosing, pressure or time with its phase and size.

  4. 4

    Across many batches it shows which factors really explain yield, quality and batch time.

  5. 5

    Process engineering and quality review the result and release new setpoints.

What changes

  • More consistent batches with less correcting and rework

  • Proven causes instead of guesses when yield varies

  • The best way to run a batch becomes the rule for every shift

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.

Assumption, editable

Your numbers

Result

+162 hfreed per year

Assumptions and calculation

demi lays each batch over the best ones and shows what differs; you decide on the new setpoint.

90 min → 63 min per case

How does it run today?

What is already digital saves less: the factor lowers the share demi brings.

How we calculate

30 batch comparisons × 90 min × 30% × 1.0 (manual) × 12 months = 9,720 min = 162 h

Value per year, before demi costs

Freed per year+162 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

+$10,530

gained per year

162 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 package

Questions about this use case

What is the golden batch, and who defines it?

The golden batch is the curve of your best batches: good yield, lab results within specification, short batch time, no correcting. The agent proposes the reference batches by these criteria; your process engineers confirm them. When the recipe or the plant changes, the reference is determined again.

Where do the process curves of the batches come from?

From the archive of your process control system, from the MES and batch records, as an export or through an interface; how they get to demi, we clarify in conversation. Lab results come from the lab system, raw material lots and quantities from the ERP. demi does not intervene in the control system.

Can demi compare batches of different lengths?

Yes. The agent compares phase by phase, such as charging, dosing, reacting and cooling, and aligns different lengths on the timeline. You see whether a batch loses time in one particular phase or was run differently overall.

Is a pattern in the data already the cause?

No, and the agent presents it that way: it names factors that go along with poorer yield or quality, with the number of batches behind them and examples. Your process engineers check whether it is the cause, for instance with a trial; only then is a new setpoint released.

See this case with your data

In the demo we show you the workflow on your own systems.

Your configurator

No use cases yet

Your configurator is still empty

Add a use case from any card or case page and see what it brings per year, with the calculation. Free, no sign-up, no commitment: try it out and change any number.

Browse use cases