The intelligence layer for metals
Every heat, decided better.
SmeltIQ builds AI copilots that understand what is happening inside the plant, recommend what should happen next, and learn from every outcome.
0.93
The intelligence layer for metals
Steel built the industrial world.
Intelligence is rebuilding how it gets made.
Every plant already runs on signals. SmeltIQ reads them together, in the seconds that decide the heat. Select a signal to see what it contributes.
Your plant already has the data.
What it lacks is the decision.
Five streams, every heat, all of them already recorded. None of them talking to each other at the moment a decision has to be made.
Bad decisions compound.
Heat after heat.
Across thousands of heats, small decisions become significant lost value.
Assumptions
Illustrative only. Uses your inputs and editable unit costs — not a SmeltIQ claim. A pilot measures the real baseline on your grades.
Excel, fixed formulas and rule-based tools are useful for standard calculations. They do not understand live plant context or the trade-offs between quality, yield, energy, productivity and cost. Excel calculates. SmeltIQ understands context.
Where the copilots work, stage by stage.
The heat is decided twice: once when the charge is set, and again while the furnace runs. Those two moments carry most of the yield, chemistry and energy — so that is where SmeltIQ starts. From casting onward, Data Copilot keeps the outcome visible and feeds it back.
Raw materials
Scrap, DRI, ferroalloys, fluxes — by lot, price and chemistry.
Recipe CopilotCharge & recipe
Charge mix and alloy plan set against grade, target band and cost.
Recipe CopilotMelting (EAF / induction)
Power, electrodes, bed condition, melt curve and tap timing.
Recipe + FurnaceRefining (LRF / VD)
Trim additions, temperature and chemistry corrections before casting.
Recipe CopilotCasting
Chemistry is locked in here — the outcome of every upstream decision.
Data CopilotRolling & finishing
Hot and cold rolling, heat treatment, inspection.
Data CopilotFinished steel
Saleable tonnes: the yield, quality and cost you actually book.
Data CopilotWhere value is melting away today.
Recipe decisions and furnace operations are the two biggest levers on yield, quality, energy and productivity. Both are still run largely on experience and static rules.
The opportunity isn't adding AI cost savings.
It's recovering value the process is already losing.
The SmeltIQ platform
Three copilots. One industrial intelligence layer.
Three products, one operating view. Each owns a different moment in the heat, and each makes the next one better informed.
Recipe Copilot
An AI decision layer that evaluates each heat, recommends the best recipe action — additions, timing, adjustments — and learns from what actually happened.
- Higher yield, more saleable steel
- Lower alloy cost through optimised additions
- Fewer off-spec heats and corrections
- Every recommendation traceable and approved
Furnace Copilot
Industrial cameras, acoustics and process signals read by AI models trained on furnace behaviour — so abnormal conditions are pre-empted, not discovered.
- Bed, crust and electrode conditions seen live
- Predicted tap time with confidence intervals
- Lower energy per tonne, steadier furnace
- Failures predicted before they stop the shop
Data Copilot
A chat and dashboard layer over every plant system. Ask a question in plain language, get the report, the trend or the compliance pack — with sources named.
- Plant health by area, asset and shift
- Breakdown history, cost and recurrence
- Compliance packs generated, not compiled
- Predictive maintenance in one view
Built for the hardest processes in industry.
Electric arc furnaces, induction units, refining vessels and smelters — across special steel, alloys, non-ferrous and mining.
Special steel
Alloy, bearing, tool and engineering grades where chemistry bands are tight.
Stainless & alloy steel
Cr and Ni additions where a percent of over-alloying is real money.
Ferroalloys & smelters
Submerged-arc and smelting furnaces where bed and electrode condition rule.
Aluminium & non-ferrous
Melting and holding furnaces, dross and energy per tonne.
Mining & mineral processing
Roasting, calcining and smelting operations at the mine site.
Anywhere metal is melted.smelted.refined.cast.
AI with an operator in the loop. Always.
Not a black box, and not autopilot. A closed loop that the operator stays in charge of.
Understand the live heat
Grade and target chemistry, charge mix and raw materials, temperature and furnace status, and how similar heats performed before.
Recommend the best action
Which additive, how much, when — with a confidence score and the reasons behind it. The operator approves, adjusts or overrides, and the reason is coded.
Learn from what happened
Actual additions, lab results, delays, corrections and operator decisions all flow back. The model improves with every heat and stays within your fallback rules.
The missing layer in the industrial stack.
Copilots are valuable because they sit in the gap that ERP, MES and SCADA leave open: the decision itself. SmeltIQ stitches enterprise systems and floor signals into one operating view, then checks, recommends, explains and learns — every heat.
What leadership sees
Don't transform the plant. Prove one decision first.
12–15 weeks. One use case. One measurable baseline. You keep your existing know-how; we connect it to live plant systems and turn it into a controlled AI layer that improves every heat.
Data & recipe audit
Grade list, input quality, formula logic, gaps.
Model & rules setup
Chemistry prediction, constraints, fallback logic.
Operator workflow
Recommendation screen, approvals, reason codes.
Live pilot
Selected grades or route, measured against baseline.
Scale blueprint
ERP/MES rollout plan, SOPs, ROI sign-off.
A spreadsheet can copy a formula.
It cannot copy this.
Domain
Heat-specific logic for chemistry, temperature, grade routes, additives and plant constraints — built with metallurgists, not adapted from generic analytics.
Data
Every heat outcome becomes training data: actual additions, lab results, delays, corrections, operator decisions. The model is yours and gets better on your plant.
Integration
ERP, MES, LIMS, SCADA, inventory and cost data stitched into one operating view. The recommendation knows what is in stock, what it costs and what the lab just said.
Governance
Approvals, reason codes, audit trail, model confidence, fallback rules and change control. Explainable to operators, auditable for leadership.
A closed-loop operating system trained on plant behaviour, connected to live constraints, and accepted by operators shift after shift.
Algorithms can be copied.
Plant intelligence compounds.
Questions we get asked
AI for steel plants, answered plainly.
The questions plant leadership, metallurgists and IT teams ask us most often in a first conversation.
What is SmeltIQ?
SmeltIQ Private Limited is a product company building AI copilots for metallurgical plants and mining operations. It provides three products — Recipe Copilot, Furnace Copilot and Data Copilot — that sit as a decision layer between enterprise systems (ERP, MES, LIMS) and shop-floor systems (SCADA, PLC, QMS). SmeltIQ reads live plant signals, recommends the next best action, explains why, and learns from the outcome of every heat.
What is an AI copilot for a steel plant?
An AI copilot for a steel plant is decision-support software that reads live production, process, quality and equipment data and recommends a specific action to the operator — for example how much ferroalloy to add and when — with a confidence score and the reasoning behind it. Unlike automation, the operator approves, adjusts or overrides every recommendation, and every decision is recorded. Unlike a dashboard, a copilot answers "what should we do next" rather than only showing what happened.
How is this different from a BI dashboard or MES report?
ERP knows the business, MES knows production, SCADA knows the process and LIMS knows chemistry. None of them decides. Dashboards and reports show what already happened, and they show only what someone thought to chart. SmeltIQ assembles all four data sources into one live context and produces a governed recommendation at the moment a decision has to be made, with approvals, reason codes and an audit trail attached.
Which furnaces and processes does SmeltIQ support?
SmeltIQ is built for electric arc furnaces (EAF), induction furnaces, ladle refining furnaces (LRF), vacuum degassing, submerged-arc and smelting furnaces. It is used in special steel, stainless and alloy steel, ferroalloys, aluminium and non-ferrous melting, and mining and mineral processing operations including roasting, calcining and smelting.
What does SmeltIQ integrate with?
ERP (including SAP PM, PP and CO modules), MES and scheduling systems, LIMS and lab systems, SCADA, PLC and DCS, inventory systems, QMS and traceability systems, maintenance systems, and industrial camera and acoustic feeds. Emissions telemetry can be pushed to ESG reporting frameworks such as SASB, CDP, UN SDGs, ISO and SBTi.
How long does a pilot take, and what does it prove?
A SmeltIQ pilot runs 12–15 weeks in five evidence-gated steps: data and recipe audit (2 weeks), model and rules setup (2–3 weeks), operator workflow (2 weeks), live pilot on selected grades or one furnace (4–6 weeks), and a scale blueprint with ROI sign-off (2 weeks). The pilot measures results against a baseline taken from your own production records rather than a vendor estimate.
Where does the value come from?
From value the process is already losing rather than from new cost savings. Audited plants typically show 1–2% yield loss attributable to recipe and charge decisions, 0.5–1.5% of output lost to quality, rework or downgrade, 0.5–1% of material cost in avoidable alloy additions, and specific furnace energy consumption ranging from roughly 420 to 775 kWh per tonne — a spread driven by decisions rather than physics.
Will operators accept it, and who is accountable for a decision?
The operator remains accountable. SmeltIQ recommends, the operator decides, and the system learns. Every recommendation carries its reasoning, a confidence score and the number of similar heats behind it. Operators can approve, adjust or override, and overrides are never blocked — they are recorded with a reason code and become training data. When model confidence is low, the system defers to the plant's existing rules.
Is plant data secure, and where do models run?
Models and dashboards run on the SmeltIQ cloud platform with direct, secure integration to plant systems; camera and process data arrive via direct integration or secure API upload. Access follows your existing roles, and every answer is traceable to source records. Model changes are versioned, tested and released under change control like any other plant change.
Do we need to replace our existing spreadsheets and formulas?
No. The pilot keeps your existing recipe logic and know-how and connects it to live plant systems. Your formulas remain the fallback layer. What changes is that recipe and furnace decisions become a controlled, learning AI layer that is explainable to operators and auditable for leadership.
Bring a copilot to your next heat.
Start with one process. One problem. One measurable outcome.
The next generation of plants won't just collect more data.
They'll make better decisions.
Every heat, decided better.