AI risk modelling that calibrates to your own numbers, for the capital decisions where the band actually costs you. We quantify the P10 / P50 / P90 of your value chain and sit on top of the tools you already run.
Built for operators, developers, and capital providers who need risk-quantified answers, not point-estimate plans. Iron ore to lithium, established commodities and the new critical minerals.
It tells you what should happen if everything goes to plan. It doesn't tell you the probability that it will.
Haulage simulation. Process plant control. Orebody calibration. None quantifies risk across the full chain at the planning horizon.
Capital allocation across operations is treated as scenario-by-scenario, not as a probability distribution under correlated 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.
Node library spans iron ore and copper through to lithium hard rock and brine, REE separation, HPAL nickel/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.
Methodology validated against publicly reported quarterly production: a full 15-quarter disclosure history on a top-tier Australian hard-rock lithium operation, matched to within 0.1% on trailing-year production with probabilistically validated uncertainty envelopes, and re-validated per-metric figures on the only producer of separated heavy rare earths outside China. Using only data the operators have made public. Direct operator engagement tightens the model further; the public-data figures represent the floor we can demonstrate today.
Run scenarios, compare side-by-side, share with stakeholders, all in-platform. No multi-month implementation. The methodology engine remains on Copula Labs infrastructure; private cloud deployment is available for operations where data residency requires it.
Heads of asset strategy, planning, finance at producing miners.
"What's the probability we meet next year's guidance under shared commodity uncertainty across our portfolio?"
Probabilistic simulation across independent assets under correlated price and FX drivers, with chain-model parameters fitted to your historical production. P10/P50/P90 envelopes for compliance-to-plan reporting at board level.
Project directors, development VPs, CFOs at projects approaching FID.
"How do we present probabilistic NPV envelopes to ECA credit committees and FID approval?"
Probabilistic NPV envelopes with explicit Monte Carlo over commodity prices, recoveries, capex, and FX. Defendable risk quantification for export credit agency debt sizing and board investment decisions.
Project directors, heads of technical services at post-DFS projects.
"What's the probability of meeting our offtake commitments over the contract term?"
DFS uncertainty stress-testing with explicit shortfall-risk quantification. Discrete event simulation of planned and unplanned outages, mapped to offtake contract penalty exposure.
Export credit agencies, project lenders, equity capital providers.
"Can we trust the operator's NPV? What's the actual risk envelope?"
Independent third-party probabilistic envelope, benchmarked against publicly reported operator data. Risk-quantified deliverables for credit committees that go beyond deterministic financial models.
Graph-based interface with stream-type validation. Drag and drop unit operations. Connections that are not physically valid are rejected.
Orebody grade, recovery, equipment availability, commodity prices, FX. Use built-in distributions or supply your own.
Monte Carlo simulation with copula correlations. Optional optimisation layer. Outputs include P10/P50/P90 envelopes and full distribution shape per metric.
History-match chain-model parameters against your production data, or against published data where direct access is restricted. The forecast envelope tightens with each fitting cycle.
Two real Australian operations, published quarterly production series, ChainVision runs. On a top-tier hard-rock lithium operation the model matches trailing-year production within 0.1% with probabilistically validated uncertainty envelopes; on the only producer of separated heavy rare earths outside China it reproduces total REO production within 0.1% and revenue within 3.1%. Using only data the operators have made public.
The validation runs use only data the operators have published in quarterly reports. Parameters are AI-tuned per operating regime, and the model's uncertainty is validated separately: P10 to P90 envelopes are tested against every published quarter, and a fully blind test checks what the model predicts before it has seen the answer. Fit accuracy and blind prediction are always reported separately.
The world needs more minerals than the existing supply base can deliver. Easy ore bodies are gone. Grades are falling. Plants are running variable feedstock through circuits that were not designed for the chemistry. Capital allocators are pricing risk that operators struggle to quantify.
Integrated decision-support has traditionally meant a multi-year build and a large internal team. That is a decade-long programme most operators cannot justify for the decisions in front of them today.
ChainVision delivers that risk layer as a product. It can be pointed at a single asset, a full portfolio, or one capital decision in weeks rather than years, at whatever scale the decision demands.
Security is assumed, not bolted on. ChainVision is engineered to clear the controls miners and their capital partners expect.
TLS 1.3 in transit with no plaintext fallback. AES-256 at rest across database, object storage, and backups.
Database-engine row-level security, with cross-tenant isolation verified in CI on every change.
Hosted in an Australian region by default. Private-cloud deployment keeps operational data inside your own tenant.
Customer data is never used to train models or build product without explicit written consent. Audit logging on every security-relevant event.
See ChainVision applied to your portfolio. Book a walkthrough of the methodology, the product, and pricing.
Book a demo Or read about our approach →