Predictive maintenance
Replaces the component before it fails and not a shift earlier, using sensor drift, load history and failure genealogy across the fleet.
- Moves
- Unplanned downtime hours
- Twin
- Asset Twin
- Capability
- Sensor ML + survival models
Industries
03 — Manufacturing & Industrials
Throughput is the product.
Asset, plant and supply chain twins that connect OEE, yield and service margin to a single financial number — and agents that hold the gain across shifts.
North Star
Contribution margin per constrained hour
The value engine, end to end
Discover → Model → Twin → Agents → Govern → Compound
The same six moves we run everywhere, expressed in the physics of this industry.
01
Discover
Where is the true constraint?
Constraint and value map
02
Model
What drives yield, downtime and cost?
Causal process model
03
Twin
What if we re-sequence or re-spec?
Plant + network twin
04
Agents
Who acts within the shift?
Scheduling and maintenance agents
05
Govern
Is it safe and standard-compliant?
IEC 62443 + safety case
06
Compound
Does OEE hold?
Outcome contract + drift monitor
Where the value is
Value pools and their drivers
Before a twin is built, we agree where the money sits and what moves it.
Value pool
Asset availability
Driver
Unplanned downtime, MTBF
Measured as
Constrained hours recovered
Value pool
Yield & quality
Driver
Scrap, rework, first-pass yield
Measured as
Material and rework cost
Value pool
Supply chain
Driver
Inventory turns, expedite freight, OTIF
Measured as
Working capital released
Value pool
Aftermarket
Driver
First-time-fix, contract attach
Measured as
Service margin per install base unit
Use cases
Where value is already sitting in Manufacturing & Industrials
Each one is a number somebody already owns, a twin that models it, and an agent that moves it inside a defined authority. Filter by function.
9 use cases
Visual quality inspection
Catches the defect at the station that caused it instead of at final inspection, and feeds the cause back into the line settings.
- Moves
- Scrap and rework cost per unit
- Twin
- Production Twin
- Capability
- Vision models
Process parameter and yield optimisation
Finds the combination of settings that holds yield under real variation in material, ambient conditions and shift.
- Moves
- First-pass yield
- Twin
- Production Twin
- Capability
- Design of experiments + optimisation
Generative design screening
Screens design and material combinations against cost, performance and manufacturability before tooling is committed.
- Moves
- Cost per qualified design
- Twin
- Production Twin
- Capability
- Surrogate modelling
Supplier risk monitoring
Scores every supplier continuously on delivery, financial and geopolitical signal, and pre-positions the second source before a line stops.
- Moves
- Line-stop hours from supply
- Twin
- Supply Chain Twin
- Capability
- Multi-signal risk scoring
Supply and inventory rebalancing
Reroutes inventory to where it will actually sell or be consumed, and updates the resource plan when the assumption breaks — not at the next planning cycle.
- Moves
- Working capital per unit of service
- Twin
- Supply Chain Twin
- Capability
- Optimisation
Spare parts and service network planning
Positions spares against predicted failure rather than historic consumption, cutting both stockouts and dead stock.
- Moves
- Fill rate at inventory cost
- Twin
- Asset Twin
- Capability
- Forecasting + optimisation
Energy intensity per unit produced
Models where energy is consumed against output and schedules load to the cheapest, cleanest window without missing the promise date.
- Moves
- kWh per unit and cost per kWh
- Twin
- Production Twin
- Capability
- Optimisation
Site safety and near-miss detection
Reads incident reports and site imagery for the pattern behind near-misses, and closes the loop with the control that prevents them.
- Moves
- Recordable incident rate
- Twin
- Process Twin
- Capability
- Vision + NLP
Outcome twins
Twins, levers, features and the agents that run them
Each twin owns a set of levers. Each feature moves one lever. Each agent executes inside a defined authority.
Asset Twin
Predict failure and schedule intervention against production value, not calendars.
Levers
Unplanned downtimeMTTRSpares availability
Features
- — Failure prediction
- — Maintenance scheduling
- — Spares optimisation
Multi-agent layer
Maintenance Planner Agent
Sequences work orders against constrained hours.
Process / Yield Twin
Hold the recipe at the profitable setpoint under real variation.
Levers
First-pass yieldScrapEnergy per unit
Features
- — Setpoint advisory
- — Root-cause analysis
- — Golden-batch comparison
Multi-agent layer
Yield Agent
Recommends and logs setpoint moves for operator approval.
Supply Network Twin
Simulate demand, supply and logistics before committing inventory.
Levers
Inventory turnsExpedite spendOTIF
Features
- — Scenario planning
- — Allocation
- — Supplier risk
Multi-agent layer
Allocation Agent
Re-allocates supply to highest-margin demand.
Service Twin
Turn the install base into predictable aftermarket margin.
Levers
First-time-fixContract attachTruck rolls
Features
- — Dispatch optimisation
- — Parts prediction
- — Service quoting
Multi-agent layer
Dispatch Agent
Matches technician, part and window in one pass.
Unit economics
ROI calculator — per feature, per unit
Set your volume, switch features on or off, and move each impact to your own evidence. Everything is expressed per produced unit.
Units produced per year
2,400,000
Contribution per produced unit
$17
Per-unit contribution; use constrained-hour economics for capital-intensive lines.
Annual program investment
$3.5m
Asset Twin
Volume — Output from recovered uptime
3.5% — Unplanned downtime converted to sellable hours.
$1.4m
annual value
Process Twin
Unit cost — Scrap and energy
3.0% — Less variation per unit produced.
$5.1m
annual value
Supply Network Twin
Unit revenue — Revenue per unit
1.5% — Supply steered to higher-margin demand.
$3.2m
annual value
Supply Network Twin
Leakage — Logistics leakage
22.0% — Fewer premium freight events per quarter.
$3.7m
annual value
Service Twin
Unit cost — Service cost per unit
1.5% — Fewer repeat visits on the install base.
$2.6m
annual value
Value by feature
Annual value contributed by each active feature at current settings.
Value and ROI vs volume
How engineered value and ROI move as units produced per year change. The marker is your current setting.
Cumulative value vs investment
Where the curve crosses the investment line is payback.
Value mix by lever
Which lever the engineered value is actually coming from.
Baseline contribution
$40.8m
Engineered annual value
$16.0m
ROI on program
357%
Payback
2.6 months
Directional model. In an engagement every number here is replaced by your measured baseline, signed off by finance, and written into the outcome contract before any agent goes live.
Scale
How this compounds from one unit to the enterprise
01
One line
Instrument the constraint, prove hours recovered.
02
One plant
Replicate to parallel lines with shared model library.
03
Network
Cross-plant allocation and one margin number.
Integrations
Domain platforms we connect
Vendor-agnostic by design. Connectors are added per engagement — no platform lock-in.
OT & MES
- Siemens Opcenter
- Rockwell FactoryTalk
- AVEVA
- PTC ThingWorx
ERP & PLM
- SAP S/4HANA
- Oracle
- Infor
- Windchill
Supply chain
- Kinaxis
- o9
- Blue Yonder
- SAP IBP
Data & cloud
- Snowflake
- Databricks
- Azure IoT
- AWS IoT SiteWise
Regulations and standards
What we are held to in this industry
Compliance is a design input, not a review gate. Every agent action is logged, attributable and reversible.
EU AI Act
Safety-component AI in machinery: conformity and oversight.
EU Machinery Regulation
Boundary between advisory agents and safety functions.
NIS2
OT cyber resilience and incident reporting.
REACH / RoHS
Material and substance traceability in decisions.
CSRD
Energy and emissions data used in optimisation must be reportable.
Standards supported
ISA-95IEC 62443OPC UAISO 9001ISO 14224ISO 50001NIST AI RMF
Next industry
Retail & Consumer