What ChainVision does, capability by capability.
Six capabilities make up the AI risk layer: chain modelling, probabilistic simulation, model fitting against operator data, multi-asset portfolio aggregation, outage modelling for offtake exposure, and chain intelligence for capital allocation. The summary below describes what each does and why it matters.
Established commodities and the new critical minerals.
Node library spans iron ore and copper through to lithium hard rock and brine, REE separation, HPAL nickel and cobalt, vanadium, antimony, tin, and battery-grade graphite. Built for the chemistry chains that gold-and-copper-heritage tools were not designed for, without abandoning the established commodities.
- +Visual node graph editor with stream-type validation: physically invalid connections are rejected.
- +Pre-built scenario templates per commodity: spodumene, REE concentrate, HPAL, lateritic Ni/Co, vanadium redox, others.
- +Custom node types available at the Enterprise tier for proprietary processing routes.
- +Performance alerts per node surface in simulation results when activation rates exceed thresholds.
Probabilistic chain simulation, end-to-end.
With the chain built, every parameter carries its own probability distribution and uncertainty propagates through every node simultaneously. Correlations between drivers (commodity prices, FX, recoveries) are preserved across the chain rather than sampled independently per node.
- +Probability distributions over the metrics that matter: production, grade, recovery, throughput, cost, revenue, NPV.
- +Correlations between drivers preserved through the simulation rather than treated independently.
- +Sensitivity ranking surfaces which drivers contribute most to variance in each output metric.
- +Outputs are full distributions, not point estimates: P10, P50, P90, and the tail beyond.
Fit to operator production, public or private.
Chain-model parameters are tuned against historical production, either the operator's own private data when they engage directly, or against publicly reported quarterly production where direct access is restricted. The forecast envelope tightens with each fitting cycle as more actuals arrive.
- +Methodology benchmarked against publicly reported production: ±1% on a top-tier Australian hard-rock lithium operation, ±3% on the only producer of separated heavy rare earths outside China, using public data alone.
- +Direct operator engagement tightens the model further; the public-data figures represent the floor we can demonstrate today.
- +Quarterly fitting cadence is included in the Operations and Enterprise tiers.
- +Fitting confidence intervals are reported alongside parameter values, so reviewers can audit which parameters are well-determined.
Multi-asset uncertainty under shared drivers.
Aggregate across independent operational assets under shared stochastic drivers: commodity prices, FX, inflation. Track probability of meeting corporate guidance against board-approved business plans, with uncertainty propagated through every node and across every asset, not collapsed to a deterministic rollup.
- +Probabilistically coherent aggregation across assets: portfolio P10/P50/P90 properly accounts for correlation between assets.
- +Shared drivers (commodity prices, FX) move together across the portfolio rather than being sampled independently per asset.
- +Compliance-to-plan reporting at board level: what is the probability of meeting the business plan number?
- +Asset-level drill-down preserved: any portfolio output can be traced back to the underlying asset distributions.
Discrete event simulation for offtake exposure.
Planned and unplanned outages at every node, mapped to offtake contract penalty exposure. Discrete event simulation is layered over the probabilistic chain simulation to surface shortfall risk against contractual obligations, not just expected production.
- +Outage frequency, duration, and severity distributions per node, calibrated against operational history where available.
- +Mass-flow-aware buffer cascades through the chain when an upstream node goes down.
- +Offtake contracts modelled as constraints with explicit penalty structures.
- +Output: probability of meeting offtake commitments over the contract term, plus expected penalty exposure.
Capital allocation and bottleneck identification.
Across thousands of simulated futures, ChainVision identifies which nodes constrain the chain most often. Confidence-weighted constraint analysis surfaces capital priorities that are invisible to deterministic planning. Sensitivity ranking shows which drivers matter for which outputs.
- +Confidence-weighted constraints: nodes ranked by how often they bind the chain across simulated futures.
- +Sensitivity ranking per output metric: which drivers contribute most to variance in production, recovery, NPV.
- +Scenario comparison side-by-side: quantified delta in P10/P50/P90 between proposed plans.
- +Outputs framed for planning conversations, not statistical artefacts that require translation.
What ChainVision does that other categories don't.
The comparison below is against generic categories of mining-decision tools, not against any specific competing product. The point is to show where the AI risk layer sits in the broader landscape.
| Capability | ChainVision | Mine planning tools | Process simulators | Strategic consulting |
|---|---|---|---|---|
| Scope | ||||
| Full chain modelled together (extraction to product sale) | + | Partial | — | Bespoke |
| Critical-minerals chemistry chains (REE, HPAL, leach) | + | — | Per unit op | Bespoke |
| Multi-asset portfolio aggregation | + | — | — | Bespoke |
| Probabilistic methods | ||||
| Probability distributions over outputs (not point estimates) | + | — | Per node | Sometimes |
| Correlations preserved between drivers | + | — | — | — |
| Uncertainty propagated end-to-end through the chain | + | — | — | — |
| Discrete event simulation for outages | + | — | Per node | — |
| Calibration & validation | ||||
| Model fitting against operator actuals | + | — | Per unit op | Bespoke |
| Methodology benchmarked against published data | + | — | — | — |
| Form factor | ||||
| Self-service product, not a consulting engagement | + | + | + | — |
| Outputs in the language planners and executives use | + | Partial | — | + |
| Methodology audit trail | + | Limited | Limited | Bespoke |
No specific competing products are named; the categories above are deliberately generic. ChainVision's positioning is not "better than tool X at task Y" but rather "the AI risk layer that sits across all of these categories without replacing the tools the planners already use day-to-day."
See it on your operation.
Book a demo to see ChainVision fitted against your own operation.
Book a demo