Agentic AI for mineral processing and plant-wide operations.
domain layers, reconciled as one
of concentrator OPEX sits in comminution
benchmark cost of an hour of unplanned downtime
The challenge
Circuits are optimised in isolation, so a gain in one place becomes a loss in another: more recovery for more reagent, a faster mill for a broken water balance, an alert that arrives after the shift it was meant to protect.
The solution
PreAur is agentic, multi-layer AI and ML that runs every circuit of a processing plant and brings them together in one control room. Ten domain layers (each a specialised agent with its own models, KPIs and conservation limits) sit on a deterministic engine floor that closes mass, water and energy balance, respects the permit in force and carries a cost. Above them, one supervisor resolves conflicts between layers and records why.
Capabilities
Grinding energy: 5–10% reductions have been documented in independent industrial deployments; this is not yet a PreAur result. Recovery, reagent, downtime and fresh-water figures are design targets that require site-specific validation.
Domain layers
Each layer is a specialised agent with its own models, KPIs and conservation limits, reconciled by one supervisor.
The AI recommends; the operator acts.
The AI puts setpoints in front of the operator for one-click approval.
The AI controls inside a signed envelope, with automatic fallback.
Illustrative visuals created for Banker Group. They convey the operating environment and are not photographs of client facilities.
Who it is for
Concentrators and refineries bringing AI agents into running operations
Sugar, ethanol and bioenergy mills
Process plants under environmental and permit pressure
Operations teams that need advice they can audit
Portfolio
Tell us where the value is. We bring the technology, the partners and the path to market.
Confidential material
Venture one-pagers are confidential and shared on request. Leave your details and we will send it to you.
