# AI Digital Twin

> AI digital twins from Boltcliq: operational twins built on your ontology and live data, with simulation and continuous sync, for plants, networks, farms, ports and cities in Africa.

Canonical URL: https://www.boltcliq.com/solutions/ai-digital-twin
Part of: https://www.boltcliq.com/llms.txt

A living model of your operation that you can query, simulate and optimize.
- A twin is only useful while it stays true to reality. We build twins on your ontology, fed by live operational data, so leaders can ask what is happening, what happens next and what to do.
- Ontology based: The twin is a model of objects, links and actions
- Live data: Continuous sync from operational systems
- Simulation: Test decisions before committing money or risk
- One picture: Teams, models and agents work from the same twin
### What we do
- The word digital twin is used for many things, from a 3D visual to a spreadsheet. We use it for something specific. A twin is a connected model of your operation: the assets, processes, people, orders and rules, with the relationships between them, kept in step with reality by live data. You can look at it, query it and run what-if scenarios against it.
- We build twins on the ontology. A pump is not just a tag in a historian. It is an object linked to the plant it serves, the work orders against it, the technician assigned, the spare parts in stock and the downstream process that depends on it. When that structure exists, a question such as which customers are affected if this pump fails has an answer, and an agent can find it without a hand-built report.
- We start narrow. One plant, one corridor, one cooperative, one network segment. The first twin answers a few high-value questions with real data. We then widen it object by object, so the twin grows with trust. Simulation arrives when the model is accurate enough that its predictions have been checked against what actually happened.
### A shared picture of the operation
- In most organizations each team holds a partial view. Operations has the schedule, maintenance has the work orders, finance has the cost and planning has the forecast. They disagree because they use different definitions and different timing.
- The twin gives them one place to look. People can zoom from the national network to a single asset and back, and every view comes from the same objects. Meetings change from arguing about whose number is right to deciding what to do.
- Views from a whole network down to a single asset
- Definitions agreed once and reused
- Models and agents read the same twin as people
### Capabilities
**Operational twins**
- We model assets, processes, locations, people and orders as connected objects. Properties come from sensors and systems, and links describe how things depend on each other. The result is a graph you can query in plain language and through APIs.
- Objects, properties and links
- Hierarchies from site to component
- Plain-language and API access
- Dependencies are where the surprises live.
**Continuous sync**
- We connect the twin to SCADA, ERP, maintenance, GPS, IoT and spreadsheets, and agree how fresh each feed must be. Each property carries a timestamp and a source, so users know how current a value is. When a feed fails, the twin says so instead of showing stale data as fact.
- Connectors to operational systems
- Freshness and source on every value
- Quality checks and gap alerts
- A twin that lies once will not be trusted again.
**Simulation and what-if**
- We build simulations on top of the twin for questions such as what if this line stops, what if demand rises by a quarter, what if a road closes for a week. Models are calibrated against history and their forecast error is reported. Scenarios can be saved, compared and shared.
- Calibrated against history
- Forecast error reported openly
- Saved and shareable scenarios
- Try the expensive decision on the model first.
**Prediction and optimisation**
- We add models for failure, demand, delay and loss, and optimisation for scheduling, routing and stock. Results are written back onto the objects, so a prediction appears beside the asset it concerns. Planners can accept a proposal or change a constraint and recompute.
- Failure and demand models
- Scheduling and routing optimisation
- Predictions attached to objects
- Insight is useful where the decision is made.
**Agents on the twin**
- Agents can read the twin and propose actions, such as raising a work order, rebooking a shipment or alerting a manager. Each action runs through the permissions and approvals for that object. A planner reviews and approves the actions that change the real world.
- Agents query objects, not raw tables
- Actions governed by permission
- Approval before real-world change
- Agents are safer when they act on a model with rules.
**Spatial and 3D views**
- Where location matters, we add maps, floor plans and 3D views over the same objects. A user can click a building, a vehicle or a field block and see its state, history and open actions. Visuals are optional and never replace the underlying model.
- Maps, plans and 3D views
- Click through to object detail
- Same data as every other view
- Some questions are easiest to answer by looking.
### Where it applies
- Energy and utilities
- A grid operator wants to know which feeders are close to overload and which customers lose supply if a transformer fails.
- The twin links transformers, feeders, meters and crews. Simulation shows the effect of a fault or a switching plan.
- The control engineer chooses the switching plan and dispatches crews.
- Mining and heavy industry
- A mine plans haulage, crushing and maintenance across shifts, but data sits in several unconnected systems.
- The twin connects fleet, plant and stockpile data. Planners test shift plans and see the effect of a breakdown.
- The mine planner approves the plan. Safety leads approve any change to operating rules.
- Ports and corridors
- Vessels, yards, gates and trucks interact, and a delay at one point cascades through the corridor.
- A corridor twin tracks containers, equipment and bookings, and simulates the effect of a berth delay or a customs backlog.
- The port operations manager reallocates resources and communicates with shipping lines.
- Agriculture value chains
- A processor depends on supply from many smallholders with variable timing and quality.
- The twin links farms, fields, deliveries, storage and factory capacity. It forecasts supply and tests transport and storage plans.
- The supply planner commits volumes and decides where to invest in storage.
### Built on
**Zeus**
- Holds the ontology of the twin: the objects, links, rules and actions, with versioning and review.
- Boltcliq product
- One definition of every asset and process.
**Argus**
- Runs ingestion, simulation and agent workloads and stores operational data on Postgres, portable across regions.
- Boltcliq product
- Compute and data can stay in-country.
**Cyclopes**
- Monitors data freshness, model error and decisions, so the twin stays trustworthy.
- Boltcliq product
- You know when the twin is wrong.
### What we do and do not do
- Model your operation as objects, links and actions
- Connect and quality-check live data feeds
- Build and calibrate simulations with your engineers
- Add maps and 3D views where useful
- Train your team to extend the twin
- Sell a 3D visual as a twin
- Show stale data as current
- Let simulation outputs act without review
- Claim accuracy before calibration
- Model everything at once
### Questions we hear
- How is this different from a 3D model?
- A 3D model shows shape. A twin holds the state, relationships and rules of the operation and updates with live data. A 3D view can sit on top, but it is optional.
- What data do we need?
- Enough to answer the first questions: asset registers, work orders, sensor or system readings and a location. We start with what exists and improve quality as we go.
- How current is the data?
- It depends on the source. Each value carries a timestamp and source, and we agree freshness targets per feed. Where a feed is slow or offline, the twin says so.
- How accurate are the simulations?
- We calibrate them against your history and report the forecast error. We do not promise accuracy before we have tested it on your data.
- Can the twin run on premises?
- Yes. Argus is designed to be portable, and we can deploy in your data centre or an in-country provider. Confirm data location rules with counsel.