Connect. Weeks, not months. Signalise connects to your historian, DCS and APC over PI and OPC UA, on dedicated, isolated edge infrastructure. Every signal is mapped onto your flowsheet and checked for quality on arrival, with every link built for audit against recognised OT and AI security standards. In Connect, your team defines the objective: how Signalise weighs throughput, recovery, reagent consumption and energy use at your plant.
Plant-wide setpoint optimisation
Every layer holds a margin.
Throughput, recovery, reagent and energy pull against each other. Signalise weighs them every minute, pushing rate when the ore allows, recovery when it pays, and cutting cyanide, lime and power when the ore does not need them, toward the objective your team sets.
Signalise reclaims it. Continuous, plant-wide setpoint optimisation lets a gold plant reclaim the cost of its hidden constraints as the ore changes.
Gold plants are among the most complex machines in industry: nonlinear, coupled and shifting with every change in ore. Your control systems, and the people who run them, capture much of that value. But fixed logic and models tuned around an operating point, refreshed by periodic step tests and studies, mean every layer holds a safe margin, because each sees only its own part of the plant. By design, and at a cost. Closing that gap needs software that learns how the plant behaves and acts on what happens next, which advances in AI and computing have only recently made possible. Signalise is that AI agent: it adjusts setpoints across the plant, every minute, toward the objectives and within the limits your team sets.
What a point is worth.
Here is what 1% more throughput and 1 point more recovery are worth over a year.
A gold plant at5.0Mtpa
- 1% more throughput
- 1,770 oz
- 1 point more recovery
- 1,930 oz
- Together, a year
- 3,700 oz
- Extra revenue
- A$22.2M
An illustrative plant at 1.2 g/t, 92% recovery and A$6,000 an ounce.
Status
Where we are
- Product
- Two years of dedicated R&D, combining the two frontiers of AI for gold processing. AutoPilot is autonomy, not a language model: trained end to end on your plant, it drives setpoints through your control systems every minute and improves the longer it runs. Analyst puts frontier language models to work, with skills, over the same data. Both on one sovereign platform, running on your stack.
- Operations
- Our agents monitor every customer model around the clock, with a Signalise AI engineer on call.
- Standards
- Built against IEC 62443, with SOC 2 Type 2 and ISO/IEC 42001 on our audit roadmap.
Built in Perth by a team that has run production data systems since the late 1990s, specialised in machine learning for mineral processing since the early 2010s, and earned control authority in live leach circuits. Meet the team
Who we serve
Built for everyone who runs a gold plant.
Every role feels the ore change differently. The operator feels it first, the GM hears about it last.
Operators
Fewer surprises on the shift.
You set up for the hardest ore you might see, because a trip costs you the shift.
Signalise moves setpoints as the ore changes, inside your limits. One click returns control.
Metallurgists
Setpoints linked to the assays.
The assay arrives hours after the feed did, and much of the week goes on gathering data.
Signalise learns which setpoints drive recovery and reagent use, with step tests you nominate. Your time goes on the metallurgy.
Processing managers
Fewer trade-offs between circuits.
Each circuit is controlled and tuned on its own, so the trade-offs between them are made by hand.
Grinding and leaching are optimised together, with the binding constraint tracked every shift.
General managers
Ounces you can audit.
A shortfall is explained after the quarter, and every fix starts with a capital submission.
Every plan is reconciled to your assays against your own baseline. No new process equipment, no shutdown.
Group and corporate
One plant first, then every operation.
Today what one site learns about its circuit stays at that site, and the next deployment starts from nothing.
The same governed platform is designed to deploy across sites in parallel, on dedicated, isolated infrastructure built for audit.
Platform
Two forms of AI. One objective.
Setpoint optimisation and plant analysis, as a service, working together for more throughput and recovery, with less reagent and energy.
AutoPilot
Acts. Autonomous setpoint optimisation through your DCS and APC, every minute, inside the limits your team sets. You choose the objective across throughput, recovery, reagent and energy. One click returns control.
Signalise platform
Analyst
Advises. A private AI assistant for your plant, with skills for constraint analysis, diagnostics, root cause analysis and simulation. Ask it anything, as often as you like. It never dispatches a setpoint.
AutoPilot, autonomous setpoint optimisation. It plans and dispatches setpoints through your existing control systems, inside their interlocks and the limits your metallurgists and operators set. AI sovereignty: models learned from your plant, running on your stack. Both on a single harness, secure and safe. Same plant. New software. More gold.
Govern. Your limits, three layers deep. Your control system stays in charge: interlocks, failsafes and HH/LL limits are governed by your operating procedures, and Signalise can never change them. Inside them, your metallurgists and operators set a range for every setpoint and weight the objectives it works toward: throughput, recovery, reagent and energy. Signalise holds itself to account, checking every simulation, prediction and plan against what the plant actually did.
- Your DCS and APC: interlocks, failsafes and HH/LL limits, never changed by Signalise
- Your ranges and objectives: a limit on every setpoint and a weighting across throughput, recovery, reagent and energy, set by your metallurgists and operators
- Signalise checks itself: every simulation, prediction and plan held close to the plant
Set limits. Every setpoint gets a range, inside the one your control system already holds.
Evaluator. Every cycle scores itself against the plant: simulation, prediction, plan, reconciliation and dispatch.
Optimise. Shadow first, then AutoPilot. Signalise learns how your circuits respond as the ore changes, then plans setpoints across grinding and leaching every minute. In Shadow it recommends and forecasts the expected uplift. When your site switches to AutoPilot, it dispatches through your DCS and APC, inside all three layers. In AutoPilot, operators can nominate setpoints for small, bounded step tests, so the plant learns which moves truly drive recovery, throughput, reagent use and energy. Every decision is evaluated against the plant and learned from, so it keeps learning the longer it runs.
Shadow
AutoPilot
Simulated demonstration plant.
Attribute. Every ounce accounted for. Every dispatched setpoint is weighed against a baseline built from your own operating history and reconciled to the assays. Feed changes, liner changes and pump swaps are identified and accounted for, so you see what Signalise contributed and what else did. The baseline itself is rechecked against recent history, continuously.
Analyse. Less time chasing data. Analyst, a private AI assistant for your gold plant. Ask it anything about the plant and it answers in seconds: what is the binding constraint right now, why did leach recovery drop overnight, where is cyanide being overdosed. It runs skills such as constraint analysis and root cause analysis, and your plant data stays sovereign, on your stack. Advisory only: it can never dispatch a setpoint.
Skills. Constraint reviews, diagnostics and event reviews, run against any circuit.
Schedules. The same checks every morning, answered before the shift starts.
Operations
Operated by Signalise. Around the clock.
Signalise is a service, not a software licence. From Perth, Signalise AI engineers work on every customer's models continuously: tracking constraints as they move, tightening the margin each layer holds, and checking every plan against what the plant actually did. Our agents monitor every site 24/7, with an AI engineer on call to your team.
Trust
Safely, and without a shutdown.
Guardrails are what make autonomy possible. A plant with firm limits can be optimised continuously.
- Standards
- Built against IEC 62443, with SOC 2 Type 2 and ISO/IEC 42001 on our audit roadmap, on a dedicated, segregated link to the OT network.
- Your limits
- Setpoints only, inside the ranges your metallurgists and operators set. Interlocks and failsafes are untouched.
- Read only until you engage
- Connect and Shadow are read only end to end. Write access happens only in AutoPilot, and only when the site turns it on. Stepping back takes one click.
- Every minute, within them
- Inside those bounds it keeps working, every minute, each plan evaluated against what the plant actually did.
FAQ
Before you ask
Start with one plant. Scale to every operation.
A three month pilot takes one plant from Connect to reconciled ounces. Connect and Shadow are read only, so you see the forecast uplift before anything is dispatched. AutoPilot runs only when your site switches it on, and the result is reconciled to your assays.
Built in Perth by specialists in gold processing data and control. Meet the team