# Why metals need different AI than generic workflow tools



> Paper mill certs and poor cut plans cost time and scrap, learn how plant-focused AI reads certs, reclaims offcuts and speeds replanning.
> 
> **URL:** https://www.gosmarter.ai/blog/metals-ai-vs-generic-workflow-tools/

**Date:** 2026-08-26
**Author:** Steph Locke

**Categories:** blog

**Tags:** artificial-intelligence, compliance, data-strategy, inventory, manufacturing, quality

## 



**Your plant does not need another bot that reads emails and moves boxes on a workflow chart.** You need [Artificial Intelligence (AI)](/hubs/metals-manufacturing-glossary/#ai-artificial-intelligence) that [reads mill certs](/hubs/mill-cert-automation/), links heat numbers to stock, cuts scrap, and helps you re-plan when the day goes sideways.

The pain is blunt: **too much typing, too much scrap, and too many bad calls made with half the facts**.

I see the split like this. Generic workflow tools sort admin. [GoSmarter](/), built by [Nightingale HQ](https://nightingalehq.ai/), sorts **metals work**. It reads certs that normal [Optical Character Recognition (OCR)](/hubs/metals-manufacturing-glossary/#ocr-optical-character-recognition) mangles. It tracks heat-level traceability. It builds cutting plans from live stock. It helps planners deal with machine downtime, rush orders, and stock limits without breaking compliance.

What you get is simple:

-   **Fewer cert errors** from manual rekeying
-   **Lower scrap** from [better cut planning and offcut reuse](https://www.gosmarter.ai/products/cutting-optimiser/)
-   **Cleaner traceability** from goods-in to dispatch
-   **Faster replanning** when machines fail or orders change
-   **Less faff** in Excel, email chains, and side spreadsheets

The numbers in the article make the point fast. Teams have cut planning time from **5-7 days to about 1 hour**. Scrap trials cut waste from **around 5% to 2.5%**. Even a **3% yield gain** on `£5 million-£10 million` of processed plate can mean **£150,000-£300,000** back in your pocket instead of in the skip.

If your software can only read a PDF but cannot tell you whether that heat can go on that job, it is not plant AI. It is office software wearing a hard hat. **That is the whole point.**

Here's how to fix it.

{{< image src="6a8e2ee4f0ae24ed42a3319f-1787706778384.jpg" alt="Generic Workflow Tools vs Metals-Specific AI: Key Differences & ROI" >}}

## Where Generic Workflow Tools Waste Your Time

### They Can Read a PDF, But Not a Mill Certificate

Generic document AI treats a mill certificate like any other PDF. That's the first problem. It might pull text off the page, but it often can't **reliably match a heat number to the actual bar or coil** it belongs to. It also falls over on chemistry tables with multi-row layouts, superscripts, footnotes, and unit swaps like MPa and N/mm2. Worse, it doesn't know that an [EN 10204](https://de.wikipedia.org/wiki/EN_10204) 3.1 certificate carries a different level of proof from a 2.2 report. If your job depends on checked material compliance, that difference matters a lot. [\[2\]](https://drametal.com/blog/material-test-reports-traceability/)[\[3\]](https://www.forgepointengineering.com/articles/mill-certificates-en-10204/)[\[4\]](https://www.testcert.co/en-GB/guides/what-is-a-heat-number)

UK service centres and stockholders get certs from mills across Europe and Asia. Templates vary. Layouts vary. Sometimes the language varies too. When OCR trips over that mess, your quality and goods-in teams pick up the slack. They check certs by hand. They retype key fields into ERP or spreadsheets. Then month-end hits, people get tired, and **mistakes creep in where they always do: during boring admin work**. [\[2\]](https://drametal.com/blog/material-test-reports-traceability/)[\[4\]](https://www.testcert.co/en-GB/guides/what-is-a-heat-number)

The result is simple and ugly. **The wrong material gets tied to the wrong job.** Your [traceability record](https://www.gosmarter.ai/hubs/integrated-cert-traceability/) breaks before the stock even moves.

Metals-specific AI handles this better because it knows what it's looking at. It recognises common EN steel designations, standard property blocks, and heat and cast number formats you see every day. It pulls structured fields such as:

-   grade
-   standard
-   heat
-   cast
-   chemistry
-   mechanical properties
-   cert type

Then it pushes that data straight into stock and quality systems, without the usual retyping circus. [\[2\]](https://drametal.com/blog/material-test-reports-traceability/)[\[3\]](https://www.forgepointengineering.com/articles/mill-certificates-en-10204/)

The same problem shows up again in cutting and scheduling. In metals, AI has to understand **stock, machines, and compliance at the same time**.

### They Track Tasks, But They Cannot Cut Steel

Plate nesting means fitting parts onto plate with as little waste as possible. Sounds simple until you get into the real job. You have to account for plate size, part shape, kerf width, minimum web size, machine bed size, grain direction, and surface finish needs. [\[5\]](https://essay.utwente.nl/essays/102916)[\[7\]](http://essay.utwente.nl/102916/) Coil allocation has its own mess: coil width, residual length, gauge tolerance, offcut reuse rules, and the need to avoid useless narrow remnants that clog the warehouse and trap cash on the rack. [\[6\]](https://resources.utec.co/plasma-cutting/plasma-cutting-nesting-material-optimization/) Bar cutting has to respect length tolerances, bundle rules, and which heats can go to which orders. That's not task tracking. That's production maths with consequences.

Manual nesting and rough-cut rules often leave **3-8% of yield on the table**, especially on mixed-thickness or multi-part jobs. A metals-specific AI engine can sort through hundreds of parts in seconds and recover material that would otherwise go straight in the scrap skip. [\[5\]](https://essay.utwente.nl/essays/102916)[\[6\]](https://resources.utec.co/plasma-cutting/plasma-cutting-nesting-material-optimization/)

On a UK operation processing **GBP5-10 million** of plate and profile each year, a 3% yield gain means **GBP150,000-GBP300,000** more product from the same tonnage. Or, if you prefer the blunt version, less scrap and lower waste disposal costs. [\[6\]](https://resources.utec.co/plasma-cutting/plasma-cutting-nesting-material-optimization/) That's the line between software that tracks work and software that helps you protect margin.

That is the minimum bar for AI in a metals plant: **read certs, cut waste, and re-plan live work**. The next section sets out what that takes.

## RMFG Factory Tour: Automating Sheet Metal Operations with Vision AI

{{< youtube width="480" height="270" layout="responsive" id="W00HFV6bRKA" >}}

## What AI Must Do to Work in a Real Metals Plant

In a metals plant, AI has three jobs. **Read certs. Cut waste. Re-plan live work.** Start with certs. Then fix yield. Then sort the schedule when the day falls apart.

### Read Certs, Link Stock and Keep the Audit Trail Clean

At goods-in, quality and stock allocation live or die on getting cert data right first time. Mill certificates turn up as scans, stamped PDFs, and multi-page reports with messy layouts. That is the usual faff.

GoSmarter, built by Nightingale HQ, reads the lot and pulls out the fields you need: heat number, cast number, grade, standard such as [EN 10025](https://landingpage.bsigroup.com/LandingPage/Series?UPI=BS%20EN%2010025) or BS EN 10204 3.1, yield strength, tensile strength, elongation, chemical composition, and product dimensions. It then links those fields to the right coil, plate, bar, or bundle in your stock system. No retyping into ERP. No side spreadsheet that one person understands. GoSmarter's [MillCert Reader](https://www.gosmarter.ai/products/millcert-reader/) saves **120+ hours a year per role** by stripping out manual retyping and filing. [\[9\]](https://www.gosmarter.ai/gosmarter-for-manufacturers/)

Every cut, process, shot-blast, and despatch updates the heat-level traceability record. A quality engineer can type in a delivery note number and see the answer straight away:

-   which heats were used
-   which test values applied
-   which certificates matter

That means audit prep stops being a multi-day paper chase. You can finish it in hours. [\[9\]](https://www.gosmarter.ai/gosmarter-for-manufacturers/)

Once traceability is clean, the next job is simple: **stop waste before it starts**.

### Build Better Cutting Plans Before Scrap Piles Up

This is where AI has to earn its keep on the shop floor. In a two-week trial, [Midland Steel](https://midlandsteelreinforcement.com/) cut scrap from about **5% to 2.5%** across **734 tonnes** and **193 jobs**. [\[10\]](https://www.gosmarter.ai/hubs/cutting-optimiser/)[\[8\]](https://www.gosmarter.ai/hubs/scrap-waste-yield-optimisation/) That result pushed the business to keep the Offcut Tracker and Scrap Weight Tracker tools for good, with a target of holding rebar and long-product scrap below **2.5%** against an industry average of **5%+%**. [\[10\]](https://www.gosmarter.ai/hubs/cutting-optimiser/)

Offcuts do not vanish into a black hole. The system treats them as stock. When you cut a bar, it records the remnant's dimensions, grade, and heat, then offers it up for future cutting plans. Production managers can see what is actually usable on the rack before they touch fresh stock. Sales teams can see real material availability before they promise lead times they cannot hit.

Then the problem changes. It is no longer about yield. It is about disruption.

### Re-plan Production Fast When the Day Goes Wrong

A saw breaks at 09:00. A rush order lands at 10:30. A heat goes on quality hold at 11:00. Generic scheduling tools treat that like a diary issue. It is not. It is a factory problem.

GoSmarter's [Production Planner](https://www.gosmarter.ai/hubs/shop-floor-planning-software/) takes live data from machine status, current work-in-progress, and updated delivery dates. Then it tests schedule options against the stuff that actually blocks work:

-   machine capacity
-   changeover times
-   crane availability
-   maximum bundle weights
-   whether the right grades and heats are available for each order

Planners get options they can check in minutes, not hours. That beats another useless alert on a screen.

MillCert Reader clears the cert bottleneck at goods-in. Cutting Plans and the Offcut Tracker claw material back before it hits the scrap skip. The Production Planner keeps the schedule honest when the day goes sideways.

The next question is how to add those tools without ripping up the systems your team already uses.

## Why [GoSmarter](https://www.gosmarter.ai/) Fits How Metals Teams Work

{{< image src="50848b877f173b9d09af2c8e83d35dcf.jpg" alt="GoSmarter" >}}

Most metals plants already run ERP, MRP, MES and a pile of spreadsheets. Software that demands a full rip-and-replace job never makes it to the shop floor. The win only shows up when the software fits what you already use.

### It Sits on Top of Your Existing Systems Without an IT Project

GoSmarter, built by Nightingale HQ, connects through CSV imports and exports or a REST API. Your current ERP, MRP or MES does not need to change [\[1\]](https://www.gosmarter.ai/hubs/integrated-planning-materials-alignment/). You can start with clean stock and cert data in Metals Manager and MillCert Reader. Then planning and inventory sync from that live record. That same record feeds the modules that sort out traceability, scrap and planning.

### Each Product Maps to a Real Shop-Floor Problem

GoSmarter fits metals plants because it works from live stock, cert and production data. Not generic task lists. Each module tackles one of the three problems already covered: traceability, yield and scheduling.

| GoSmarter Product | Shop-Floor Problem | Annual price |
| --- | --- | --- |
| [GoSmarter Insights](https://www.gosmarter.ai/docs/track-scrap/) | Scrap weight, cost and carbon calculations | Free |
| [Product Lineage](https://www.gosmarter.ai/features/) | Mill cert scanning, heat-code linking, traceability | £275 / month |
| [Business Manager](https://www.gosmarter.ai/features/metals-manager/) | Inventory tracking, order management and scrap tracking | £400 / month |
| Production Planner | Long-product cutting plans against live inventory and orders | Price on application |

That means you can go after the bottleneck that hurts margin first, instead of buying a giant software bundle and hoping for the best.

Product Lineage handles the [cert and traceability work](https://www.gosmarter.ai/blog/gosmarter-vs-generic-ocr-mill-cert/) covered earlier in this article. It turns cert handling into a live stock record.

Business Manager covers inventory and order control.

Production Planner builds [long-product cutting plans](https://www.gosmarter.ai/docs/what-is-cutting-optimisation/) from live stock and orders. It is built for long-product cutting. Not generic 2D nesting.

Plants can start with one module and add others later. Metals AI has to work with live materials and live constraints. That is the line between software that helps and software that just adds more faff.

## Test GoSmarter on Your Next Batch of Mill Certs

If **paper certs** still clog the works, start there. Pull 20-50 live mill certificates from a few mills. Then upload them to **GoSmarter MillCert Reader**, built by Nightingale HQ, instead of typing the lot into Excel or your ERP.

MillCert Reader pulls out the fields you need and links them to stock records. Run it alongside your current process for one week. That shows up **data errors** and **broken links** fast. Once you clean up the cert data, move to the next choke point.

Use **Product Lineage** when traceability packs eat hours. That matters for structural plate, safety-critical profiles, and any job that needs [EN 10204 Type 3.1](https://www.gosmarter.ai/docs/what-is-en-10204/) traceability. [\[13\]](https://www.mckinsey.com/~/media/McKinsey/Industries/Metals%20and%20Mining/Our%20Insights/How%20a%20steel%20company%20embraced%20digital%20disruption/How-a-steel-company-embraced-digital-disruption.pdf)

If traceability already behaves itself, go after scrap. Use **Production Planner** when **scrap and offcuts** chew through margin. Midland Steel ran a two-week trial across 734 tonnes of rebar and 193 jobs. They recorded a 50% cut in scrap, from about 5% to 2.5%. [\[8\]](https://www.gosmarter.ai/hubs/scrap-waste-yield-optimisation/)[\[11\]](https://www.gosmarter.ai/hubs/roi-ai-metals-manufacturing/)[\[12\]](https://www.gosmarter.ai/casestudies/midland-steel/) At 100 tonnes a week, that saves about 2.5 tonnes of scrap each week, or about **GBP46,800 a year** at GBP360 per tonne. [\[11\]](https://www.gosmarter.ai/hubs/roi-ai-metals-manufacturing/)

Run one pilot first. Then expand.

-   Start narrow: one product family, one site, four to six weeks.
-   Track cert processing time, data accuracy, and either traceability completeness or scrap rate.
-   Pick the metric that matches your pilot path.
-   Run the pilot on live material.
-   Review the results with operations and finance.

## FAQs

{{< faq question="How does metals AI improve quote accuracy?" >}}
Metals AI improves quote accuracy with a **specialised costing engine** that factors in **more than 50 variables**. That includes CAD specifications, machine costs, and live material prices.

It also connects to live inventory and verified real-time data. So instead of quoting from **outdated spreadsheets** and a bit of hopeful guesswork, you price work based on what's actually in the yard. It prioritises stock you already hold, which cuts quoting errors and stops you baking bad numbers into the job from the start.
{{< /faq >}}

{{< faq question="Can it work with our current ERP or MES?" >}}
Yes. GoSmarter, built by Nightingale HQ, works as an **AI overlay** on top of your current ERP or MES. It does **not** rip out the system you already have.

It reads and updates data through CSV import and export, or through a REST API with OAuth 2.0. It also syncs heat- and certificate-aware stock, so your planning matches what you **actually** have on the shop floor, in near real time.
{{< /faq >}}

{{< faq question="What should we pilot first in our plant?" >}}
Don't try to fix the whole plant in one go. That's how good ideas turn into **expensive software theatre**.

Start with the pain that's biting you now. Manual data entry. Material waste. Pick one problem. Then pilot one module, like **scrap optimisation** or **mill certificate management**.

If you're working on predictive maintenance or quality control, keep the scope tight. Start with one product family in a single diameter range. Link certificates, inspection results, and scrap records first. Build a clean, heat-linked dataset before you scale anything.
{{< /faq >}}