POL/00 AI plant intelligence assistant

POL — the AI brain for your steel plant.

Ask questions in natural language. Understand production. Find losses. Investigate abnormal heats. Get management insights — without digging through spreadsheets.

Book a POL Demo See it work

6 live

Furnaces covered

8,400+

Heats indexed

3 per reply

Answer layers

2 switchable

Response modes

POL/01 How it's structured

One layer captures the heat. Another explains it.

POL doesn't replace the production database or the live dashboard — it sits alongside them as a question-answering layer. Every response is built from three clearly separated pieces, so a shift lead always knows what's a plant fact, what's outside context, and what's POL's own call.

01 · DATA

Grounded in the plant

Pulled straight from the same heat-by-heat production database as the live dashboard and entry forms — furnace and ladle numbers, power-per-ton, tap-to-tap duration, tapping temperature — cited down to the specific heat number.

02 · INTERNET

Checked against the outside world

Looks up standard plant-floor benchmarks and known causes — refractory wear, slag buildup, scrap density, holding-time losses — kept visibly separate from what came out of your database.

03 · POL SUGGESTS

A judgement call, not just a readout

Combines the two into a plain-language recommendation — which furnace to inspect first, and why — ranked, reasoned, and specific enough to act on without translating it yourself.

POL/02 What you can ask

Every conversation starts with a nudge.

POL surfaces starter prompts tuned to what usually needs a second look — or ask your own question, narrow or broad.

Ask

"Why did production fall yesterday?"

Investigate

"Compare Furnace 1 and Furnace 2."

Analyze

"Find the heats with abnormal cycle times."

Monitor

"Which consumables are approaching reorder levels?"

Report

"Give me today's plant summary."

Watch

"Is Furnace 1A's power-per-ton creeping upward?"

POL/03 See it work

Two real questions, answered the same way every time.

Ask something narrow — a single furnace, a single shift — or something broad enough to span the whole heat history. The structure of the answer never changes, only how much of the plant it has to reason across.

POL · cross-furnace drift, 8,422 heats

Across the 8,422 heats logged, which of the 6 running furnaces is showing the worst tap-to-tap creep or power-per-ton drift over the last 48 hours?

DataOver the last 48 hours, Furnace 2B logged the highest sustained power consumption across active furnaces, averaging 872.9 kWh/t across 13 heats, peaking at 1,199.78 kWh/t. Furnace 1A recorded the single worst individual power spike, and Furnace 3B recorded the longest individual tap-to-tap duration.

InternetNo external lookup was needed here.

POL suggestsPrioritize an immediate inspection of Furnace 2B, followed by a check on 1A's electrical efficiency and 3B's turnaround delays, because 2B is consistently drifting highest above nominal power consumption targets.

POL · week-over-week comparison with outside context

Compare furnace power per ton over the last week — is any furnace drifting worse, and what typically causes that on a plant floor?

DataAcross 200 heats, plant baseline furnace power averages ~790–825 kWh/t, but Furnace 2B displays a clear worsening upward drift mid-week, while Furnace 1A holds a steady baseline punctuated by isolated holding delays.

InternetStandard steel plant operations benchmark specific energy consumption drift to progressive refractory lining erosion or slag/skull buildup, lower scrap bulk density, excessive tapping superheat, and prolonged holding times.

POL suggestsCheck Furnace 2B's refractory lining profile and slag buildup at the next inspection, and audit crane/ladle holding times on 1A's spike heats — 2B shows steady cycle-over-cycle erosion, while 1A's spikes stem from operational delays rather than furnace health drift.

POL/04 Access & modes

Two ways to ask. One place to open it.

POL lives inside the same admin panel that runs your live dashboard — one click in, one click back, no separate login and no separate screen to check.

Everyday POL chat

Fast, conversational

Built for the question that comes up mid-shift — a quick check on one furnace, one bay, one number that looks off — answered in seconds without breaking stride.

Hard analysis

Deeper reasoning, full history

For end-of-shift or investigative questions that need to reason across the entire heat log, with the underlying model swappable per conversation, without changing how POL structures an answer.

Built-in caution, by design: POL can be wrong about outside context — always double-check anything safety-critical. Plant data is cited precisely; outside context is always labelled as such.