A small scrap yard can use AI to lay out the daily price sheet from your own margin rules, check weighbridge ticket exports for odd weights and duplicates, turn them into monthly customer statements, and draft emails to buyers. It should not decide who you may buy from or how you pay: those are compliance questions for people who know your licence.
Yard software already does a lot of this work, and the sensible question is what sits around it. The useful split has three parts: what a dedicated yard system covers, where a chat assistant and a spreadsheet fill the gaps, and what always stays with a named person at the gate or the desk.
What yard software covers, and what falls outside it
Scrap yard platforms are built around the scales. ReSpark's pages describe automated scale workflows, inbound ticketing, ID scanning, price lists, compliance documentation, payments and accounting integrations, and a dispatch tool that reads requests from email and drafts tickets for one-click approval. Scrap Dragon's pages describe reading scales, capturing photos and building an audit trail, built-in compliance integrations, several payout options, inventory and sales order management, and links to common accounting packages. Both are vendor marketing pages, so treat claims such as speed at the gate as claims, and ask for a demo on your own materials.
If you already run one of these, much of what follows is available inside it, and you should use its version first, since it sits on your real records. If you run a spreadsheet, a basic scale printout or a lightweight system, the gaps look like this:
| Job | Yard software usually does | Left for you, and where AI can help |
|---|---|---|
| Reading the scale and printing the ticket | Yes | Nothing; this is hardware and a system |
| Seller ID and compliance records | Yes, in the better systems | A person checks at the gate; AI can draft an internal checklist |
| Daily price list | Stores and applies prices | Working out the prices from buyer quotes and margins, writing the sheet |
| Checking ticket exports | Some validation | Spotting odd weights, duplicates and missing fields in an export |
| Customer statements and reminders | Often in a billing module | Cover emails, summaries, polite chasers |
| Selling loads to mills and traders | Sales orders and dispatch | Offer emails, comparing replies |
The rest of this tutorial is the right-hand column.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe daily price sheet: from buyer quotes to a signed-off page
Most yards set what they pay each morning from what their own buyers, usually mills and traders, will pay them, less haulage, less a margin that depends on the material. That is arithmetic you can write down once and apply every day. The risk is not the arithmetic but the copy: a price typed by hand into a board at the gate, a sheet emailed to regular customers, and a price in the system that no longer agree.
Keep the working in a spreadsheet with one row per material: the buyer's quote per tonne, the haulage per tonne, your margin, and the price you pay. The formula is (quote less haulage) times (1 minus margin), rounded down to a figure you will say out loud. An illustrative morning, with invented figures, a flat haulage of $18 a tonne and margins that differ by material:
| Material | Buyer quote per tonne | Margin | Worked price | Sheet price (rounded down) |
|---|---|---|---|---|
| Clean copper | $7,800 | 8% | $7,159.44 | $7,155 |
| Aluminium, clean | $1,900 | 12% | $1,656.16 | $1,655 |
| Heavy steel | $290 | 20% | $217.60 | $215 |
| Stainless | $1,150 | 15% | $962.20 | $960 |
Where does AI fit? In three small ways. First, reading the buyers' emails. Quotes arrive as free text, some as attachments, some as a forwarded chat; an assistant can pull "copper 7,800, steel 290" out of a messy message so you do not retype it. Second, the sanity check against yesterday. Third, the layout of the sheet for the gate and for regular customers.
Here are yesterday's sheet prices and today's buyer quotes (from emails below).
1. Extract today's quote per material, per tonne, exactly as written. If a quote is unclear or in a different unit, write CHECK and quote the line.
2. Put the quotes in a table beside yesterday's sheet prices.
3. Show the percentage change for each material and flag any move of more than 5% either way.
4. Do NOT calculate the new sheet prices. I will do that in my spreadsheet.
Buyer emails:
[paste]Notice what the prompt refuses: the assistant does not work out your prices, because the formula lives in a spreadsheet you control, where it can be audited. After you apply the formula, ask the assistant to lay out the sign-off version: the materials and prices, the date and time, the words "valid until [time] or until replaced", and the name of the person who approved it. The final line matters. The usual failure is an old sheet still on a board or in an inbox. Why AI is poor at sums in quotes and invoices is the reason the arithmetic stays out of the assistant.
If you keep a standard buyer-facing message, such as "today's prices are attached, valid until 5 pm", the assistant can produce it in your wording each day. Review it before sending; nothing leaves the yard without a person looking at it.
Weighbridge tickets: checking an export before it becomes money
A weighbridge ticket records a vehicle's gross weight loaded, its tare weight empty, and the net weight of material, with a ticket number, date, customer, material and price. Net is gross less tare, and the amount paid is net in tonnes times the price. Errors here are errors in money, so a monthly check of the export is worth the half hour.
Export a month of tickets as a CSV or spreadsheet and treat the ticket number as text, since leading zeros and long numbers are damaged when a spreadsheet opens them as numbers. Then ask for specific checks that a person would otherwise do by eye:
This is a month of weighbridge tickets (columns: ticket, date, customer, vehicle, material, gross kg, tare kg, net kg, price per tonne, amount).
Check and list, without changing anything:
1. Any row where net does not equal gross minus tare.
2. Any row where net is zero, negative or missing.
3. Duplicate ticket numbers.
4. Any vehicle whose tare differs by more than 300 kg between tickets in the month.
5. Any row where amount does not equal net/1000 times price (allow 1 cent).
6. Any material with a price that differs from the price sheet for that date. (Price sheet attached.)
Show the ticket, the problem and the figures you used.Tolerances such as 300 kg are yours to set from your own scale's history; a vehicle's tare moves a little with fuel and what is in the cab, but there is no universal figure. If your spreadsheet can do checks 1 to 3 and 5 with formulas, let it, since a formula never mis-adds. Ask the assistant for the checks that need judgement, the vehicle tare comparison and the price-sheet matching, and for a plain-language list of what it found.
One customer's month: six tickets, one wrong tare
An illustrative month for a roofing firm that sells scrap to the yard. The six tickets below include one problem. Weights are in kilograms and prices in dollars per tonne.
| Ticket | Material | Gross | Tare | Net | Price | Amount |
|---|---|---|---|---|---|---|
| T-20417 | Heavy steel | 9,840 | 5,630 | 4,210 | $215 | $905.15 |
| T-20431 | Aluminium, clean | 7,260 | 6,240 | 1,020 | $1,655 | $1,688.10 |
| T-20458 | Heavy steel | 11,320 | 7,020 | 4,300 | $215 | $924.50 |
| T-20472 | Clean copper | 6,885 | 5,480 | 1,405 | $7,155 | $10,052.78 |
| T-20490 | Heavy steel | 10,500 | 6,100 | 4,400 | $215 | $946.00 |
| T-20503 | Aluminium, clean | 7,100 | 6,240 | 860 | $1,655 | $1,423.30 |
The six amounts add up to $15,939.83. The vehicle check, tickets T-20417 and T-20458, flags a problem. Both are heavy steel and, from the vehicle column (left out of the table to keep it narrow), the same truck, but the tare reads 5,630 kg on one and 7,020 kg on the other: a gap of 1,390 kg. The other four tickets came in on the firm's two other vehicles, whose tares stay within 300 kg of their own earlier readings. If the lower tare is right, T-20458's net is 5,690 kg and the amount owed is $1,223.35, which is $298.85 more than was recorded, and the month's total becomes $16,238.68.
The assistant cannot tell you which tare is right, and it should say so. A person looks at the ticket photos, the second weighing if there was one, and the scale log, then corrects the record or confirms it. Whether the customer is owed money, or the yard's records need a note, is a business decision. The check did its job: it showed the discrepancy before the statement went out, not after the customer queried it.
Monthly statements from the corrected export
With the tickets checked, a statement per customer is a pivot: customer, tickets, material, net tonnes and amount, with a total. A spreadsheet pivot table does it exactly. The assistant adds value in the cover note, particularly for customers with queries.
Write a short covering email to accompany this month's statement for [customer].
Facts (use only these): 6 tickets between 2 and 29 [month]; total amount $16,238.68; paid by [method] on [date]; ticket T-20458 was corrected after a tare check, increasing the amount by $298.85.
Tone: plain, polite. Explain the correction in one sentence. Invite questions. Under 120 words. Do not add or change any figure.That email is easy to get wrong if the assistant does the sums, and easy to get right when the figures arrive in the prompt. Notice that the correction is stated openly. Customers accept a corrected ticket far more readily than a quiet change. For customers who owe you money rather than the reverse, such as account customers who buy offcuts, the same pattern applies to reminders; chasing late invoices and setting credit holds describes a tone that keeps the relationship.
The ticket file is worth asking questions of too: which customers bring the most by tonnage, which material has the thinnest margin after haulage, which weekday is busiest at the scale. Six questions to ask your sales spreadsheet shows the pattern, and the same questions fit a ticket export if you remove personal details first.
Emails to buyers: offers, confirmations and comparisons
Selling a load to a mill or a trader is mostly email: what you have, how much, when it can be collected and what you want. An assistant can write the offer from a short stock list, which saves time on the tenth message of the week.
Write a short offer email to [buyer] for the following, from [yard]:
- Material: [material], approx. [X] tonnes, loose/baled, ready for collection from [date]
- Quality notes: [e.g. sorted, magnet-tested, free of contaminants]
- Photos attached: yes
- Price: [leave blank: I will add]
Ask them to reply with price, collection day and payment terms. Under 100 words. Do not state any price or quality claim I did not give.Two rules save arguments later. The quality statement is a claim you will be held to, so it must be what you actually verified; the assistant must not embroider it. And a price in an email is an offer that can be accepted, so keep prices out of the assistant's hands until you have decided.
When several buyers reply, paste the replies and ask for a comparison: price per tonne, collection date, payment terms, and anything conditional. Ask the assistant to put the conditions in a separate column, because the highest headline price often hides a deduction for moisture or contaminants. The decision remains yours. For lots that are not scrap at all but surplus stock, offers to trade buyers for excess inventory follows a similar approach.
Compliance and cash: where AI stays out of the decision
Scrap buying is regulated more than most small trades. Depending on where you operate and what your licence says, you may have to identify sellers, record what you bought and from whom, hold records for a set time, restrict cash payments, refuse certain items and report to a regulator or the police. These rules change, differ between places and carry real penalties. This tutorial does not tell you what they are, and neither should a chat assistant.
What AI can do safely is help you work with your own rules once you have them in writing:
- Turn the conditions of your licence, as supplied by the licensing body, into a one-page gate checklist in plain language, which a person then checks against the original.
- Write a polite refusal wording for a sale you cannot accept, such as one without the required identification.
- Summarise your own records for an inspection by listing, from your export, the purchases missing a required field (not whether the purchase was lawful).
- Prepare a list of questions for the licensing body or an adviser when you are unsure.
What it must not do is rule on a particular purchase, or say that your records meet a requirement. Put that boundary in writing for staff: the person at the gate decides, using the licence conditions, and escalates to the owner when unsure. Check the checklist's wording with the body that licenses you, and when you are unsure, ask them. That is also what to do about cash: the permitted methods and limits are set by the rules where you work, so confirm them there. Software systems usually build those rules into the payout step, which is a good reason to prefer one if you handle much cash.
Keep personal data out of chat tools where you can. A ticket check does not need a seller's name, address or ID number. Remove those columns before pasting, use a business plan where content is not used for training by default, and keep full records in your own system. The thinking in whether customer data is safe in ChatGPT applies directly.
A downgrade dispute, written up from the ticket trail
Buyers sometimes pay less than the quote because a load contained something that should not have been there: moisture, plastic, or a different alloy. When the deduction arrives, the yard has to decide whether to pass it on to the person who supplied the material, and the evidence is scattered across a ticket, a photo and an email. An assistant is a good tool for pulling that trail into one page.
Say a trader reports that a load of aluminium sent from the yard was downgraded by $140 a tonne on 2.4 tonnes, a $336 deduction, because of attached steel. Paste the buyer's email, your outbound ticket and the inbound tickets for that material from the previous two days (customer details removed), and ask:
From the buyer's email and these tickets, produce a one-page timeline: which inbound tickets could have contributed to this load, with date, net weight and any note. State what the records show and what they do not. Do not decide who is at fault.The result is a list, not a verdict: perhaps three inbound tickets, one with a note about attached fittings, and nothing in the records that identifies which bundle contained the steel. That is the honest answer, and it tells the owner something practical: the sorting step, not a supplier, is the weak point, and a photo at the pile before baling would help next time. Whether to claim from a supplier, accept the deduction or dispute the buyer's grading is the owner's decision, using the evidence on the page. The assistant has shortened an afternoon's search through paperwork to a few minutes, and nobody has been accused of anything the records do not support.
The price board that was a day out
An illustrative failure. A yard's copper price rose sharply overnight, and the morning's sheet, correctly worked, raised the price paid. The sheet went to regular customers by email at 8:10 but the printed board at the gate was updated at 9:30. For 80 minutes, customers who had seen the email arrived expecting the higher price and found the old one on the board. There were arguments, a refused sale and one customer who left to sell elsewhere.
The fault was not the arithmetic. The yard had two copies of a price with no rule for which was current. The fix was small and procedural: one sheet with a time of issue and "supersedes all earlier sheets", the board and the email updated in a fixed order by the same person within ten minutes, and the system price changed first. The assistant helped by producing the board text and the email from the same table, so they could not differ by a typo, but the control was the procedure.
A morning routine that fits in twenty minutes
- Quotes in. Paste buyer emails into the quote prompt, check the extraction against the originals, enter quotes in the spreadsheet.
- Prices out. Let the formula produce the sheet prices; compare with yesterday and look at any move over 5%.
- Sign off. A named person approves, adds the time, and the board, system and customer email are updated in the same order every day.
- Sell. Draft offers for loads ready to go, add prices yourself, send.
- Weekly. Export tickets, run the checks, resolve flagged rows, note the corrections.
- Monthly. Statements from the corrected export, cover emails, reminders for overdue accounts.
Run it by hand for a month before you automate any part. At the end of the month, count the corrections you made to the assistant's output, the tickets the checks caught, and the minutes saved. If you then want the sheet to flow into the yard system automatically, that is a job for the system's own import or an integration tool, and cleaning messy data is the discipline that makes the import reliable.
Questions from yard owners about AI and yard software
Do I need yard software if I use AI?
Yes, if you weigh, pay and report compliance records daily. Dedicated yard software reads the scales, prints tickets, keeps ID and payment records and often handles regulator reporting. AI does not replace those jobs. It helps with the writing, checking and summarising that sits around them.
Can AI decide whether I can buy from a seller?
No. Rules on seller identification, record keeping, payment method and restricted items depend on where you operate and what your licence says. A person who knows your licence conditions decides at the gate. Use AI to draft a checklist from your own rules, then have the licensing body or an adviser confirm it.
Is it safe to put weighbridge data into a chat assistant?
Strip personal data first. Ticket numbers, weights, materials and prices are fine for most purposes, but names, ID details, vehicle registrations and payment details are not needed for checks and statements. Use a business plan where content is not used for training by default, and keep the full records in your own system.
Who sets the daily prices?
A person does. The assistant can lay out the sheet, apply your margin rules and flag large moves, but your buyers' quotes, your costs and your appetite for stock decide what you pay. Sign the sheet off each morning with a time on it, and withdraw yesterday's copy.
Further reads
- Best AI Tool for Pricing Wholesale Goods: Chat Assistant or App? — Compare ways to keep a trade price list consistent.
- Stock Count Variance: Find Why Your Count and System Disagree — Find why a stock count and the system disagree.
- AI in Warehouse Operations: Trace Pick Errors Back to Their Cause — Trace repeated errors back to their cause, as with tare slips.
- AI in Fleet Management: Auditing Yard Time and Tolls From Telematics — Audit time on the yard and on the road from vehicle data.
- AI Email Triage for Shared Inboxes: Sales, Support, and Invoices — Sort buyer quotes, enquiries and invoices in one inbox.
- Plant Hire Software: What Small UK Yards Should Compare Before Buying — Borrow its before-you-buy questions when you trial a yard system.
- Skip Hire Software: A Buying Checklist for Quotes and Permits — Eight quote inputs, a permit tracker, a run sheet and waste paperwork, with what five skip hire vendors print on their pages and a 45-minute demo.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: the product pages of ReSpark and Scrap Dragon (yard-management software), checked in November 2026, which are vendor marketing statements, plus general scrap trade practice. Rules on identification, records and payments differ by licence and market; confirm them with the body that licenses your yard.