Discuss your project
← All use cases

Use case · Production & Maintenance

A signal.
The cause.
The right moment.

We start from your machine data, maintenance plans and failure history. Together with maintenance, production and IT we clarify which signals count, who may create work orders and when maintenance decides. The result is your workflow for predictive maintenance.

CORAProcess assistant

Your team

“Use our sensor data, the maintenance plans and the failure history. Create a workflow that detects anomalies on our equipment early: compare temperature, vibration and running times with normal behaviour. When something is unusual, search for similar past failures and the relevant sections of the manuals. Propose a maintenance slot in the next production window. Maintenance creates work orders after review.”

CORA is drafting the process …

The description becomes the workflow

Describe once.
Run every time.

Your description becomes a continuous monitoring workflow: from the sensor signal to a scheduled work order.

CORA builds the process. Your team reviews the test cases and approves it. Only then does it run for every matching event – with your permissions.

Monitor signals · Narrow down causes · Schedule maintenance

  1. Sensor data
  2. Detect anomaly
  3. Narrow down cause
  4. Maintenance proposal
  5. Work order
Discuss your project

One run in detail

How CORA detects a bearing failure.

An anomaly on a press: from the sensor signal to a commissioned bearing replacement.

All names, numbers and values in this example are fictitious.

  1. TriggerVibration on press 2 has been rising for three days.
  2. IntegrationRead data

    Sensor data of the last 30 days read: vibration, bearing temperature and running time.

  3. CodeCompare with normal behaviour
    MeasurementNormalCurrent
    Vibration2.1 mm/s3.4 mm/s
    Bearing temperature58 °C64 °C
    Running time since service–1,840 h
  4. RuleWarning level

    Warning level 2: schedule maintenance within seven days.

  5. AIFind similar cases
    A similar pattern occurred with failure ST-118: the cause was a worn main bearing. The manufacturer manual recommends a bearing replacement for this pattern (section 6.3). A spare bearing is in stock.
  6. CodeFind window

    Next suitable maintenance window: Saturday, early shift – no orders planned on press 2.

  7. HumanCommission

    Maintenance confirms the finding and commissions the bearing replacement.

  8. IntegrationCreate work order

    Work order WA-3307 created in the maintenance system, with findings, history and sources.

From existing documents to a tested workflow

From existing documents
to a tested workflow

1Your documents and example cases

DocumentMaintenance plan press 2 (fictitious)
  • Intervals
ExampleFailure ST-118 (fictitious)
  • Cause
  • Action
DocumentManufacturer manual
  • Failure patterns

2Business systems and data

SystemMachine data
  • Temperature
  • Vibration
  • Running time
SystemMaintenance
  • Work orders
  • Spare parts
RuleWarning level
  • Defines urgency

3People, roles and rights

Role cardMaintenance
Create work order
Role cardProduction
Agree window
Machine controlBlocked
No access
  1. Build process
  2. Test
  3. Approve

Running process

  1. Sensor data
  2. Check normal behaviour
Normal
  1. Keep monitoring
  2. Log
Unusual
  1. Narrow down cause
  2. Maintenance approval
Critical
  1. Shift lead at once

Process steps in detail

What AI does – and what rules, code and integrations handle.

CORA uses AI where language, context or unstructured information needs to be understood. Calculations run as tested code, limits and responsibilities as rules, system access through defined integrations. Approvals stay with people.

  • AILanguage model: understands, summarises, drafts1×
  • RuleFixed conditions, thresholds, responsibilities1×
  • CodeCalculations and models, versioned and testable2×
  • IntegrationRead and write access via MCP servers or interfaces2×
  • HumanReview, approval and decision1×
  1. 01

    Collect data

    Temperature, vibration and running times flow in continuously.

    Integration

    Implementation: Interface to machine data acquisition, read-only

  2. 02

    Compare with normal behaviour

    Deviations from usual behaviour are detected before limits are exceeded.

    Code

    Implementation: Anomaly detection per asset, trained on your histories, versioned

  3. 03

    Set warning level

    The level decides whether to monitor, schedule or notify immediately.

    Rule

    Implementation: Limits and escalation levels defined by maintenance

  4. 04

    Find similar cases

    The AI finds comparable past failures and relevant manual sections and summarises possible causes.

    AI

    Implementation: Search in company memory: failure reports and manuals

  5. 05

    Propose maintenance window

    The next suitable window is calculated, including a note on missing parts.

    Code

    Implementation: Matching against production plan and spare parts stock

  6. 06

    Review and commission

    Maintenance assesses the proposal and decides on the work order.

    Human

    Implementation: Maintenance

  7. 07

    Create work order

    After approval, the work order is created with findings and sources.

    Integration

    Implementation: Write access to work orders in the maintenance system

Connected systems

  • Machine data acquisitionreads sensor valuesvia interface
  • Maintenance systemreads history and spare parts, writes work orders after approvalvia MCP server
  • ERP / production planningreads production windowsvia MCP server
  • Document storagereads manuals and failure reportsvia MCP server
  • Company sign-inchecks identity and rolevia single sign-on

How we work

We start with your
actual work.

  1. 01

    Understand existing documents

    Maintenance plans, manuals, failure reports and sensor histories show us which patterns precede failures. From them we derive measurements, normal ranges and escalation levels. We also capture the experience of your maintenance team.

  2. 02

    Connect data and business systems

    With your IT we connect machine data, maintenance system, production planning and document storage. Asset ID and part number link the data. We define where data is only read and where draft work orders are created.

  3. 03

    Assign identities, rights and roles

    Identity comes from your company sign-in. Maintenance creates work orders, production sees the plans for its lines. There is no control access to machines; the AI acts on behalf of the responsible person.

  4. 04

    Test together and go live

    From the description, documents and clarified system access, CORA drafts the process. Together with your business unit we select test cases and expected results. Testing runs without write access. Example tests: Normal operation, slowly rising vibration, sensor failure, anomaly without matching failure history.

Your team decides on rules, AI steps and approvals. We activate the tested workflow, compare effort and results and practise the next changes in the skill workshop. Integrations, knowledge and reusable building blocks remain as the foundation for further processes.

Discuss your project

How the case
runs

CORA continuously compares sensor data with the normal behaviour of each asset. For anomalies it searches similar failures and manual sections, calculates a suitable maintenance window and presents a reasoned proposal to maintenance.

What it
produces

Warnings with level and history, likely causes with sources, a proposed maintenance window with spare parts status and the approved work order.

Where your team
decides

Maintenance reviews every proposal and creates work orders. CORA does not intervene in machine control. Critical values go to the shift lead immediately.

More use cases

We show you where AI can take work off your hands right away.

Discuss your project