ProcurementNewVerified Real deployment at a named or precisely described company, with a measured result from a primary source.

AI forecasts aluminium, copper and plastic prices so buyers know when to place large raw-material orders (La Cibek, Ormazabal)

Added 2026-10-30 · Last updated 2026-10-30

Manufacturers in this case: La Cibek · Ormazabal · all manufacturers using AI

Role

Purchasing of raw materials and commodities (metals, resins, feedstock) · Typical users: buyer or purchasing manager for raw materials, owner or general manager of a small manufacturer, finance or controlling

The problem

For many manufacturers, metals or resins are the biggest cost line, and their price can move by 10% or more in a few weeks. Selling prices are often fixed in a quote or contract long before the material is bought, so a purchase made at the wrong moment eats the margin. Most buyers decide when to buy from experience, supplier talk and a glance at the exchange price. Ormazabal, a Spanish maker of transformers, switchgear and other electrical grid equipment, buys thousands of tonnes of copper, aluminium and galvanised steel a year; once a customer contract is signed its price is locked, even if material costs rise sharply the next day, so without a forecast 'procurement becomes a gamble' across the dozens of contracts it signs every month. La Cibek, an Italian aluminium and zamak die-caster near Brescia, wanted to anticipate price increases and market swings and avoid material shortages. An Asian plastics converter had to confirm LLDPE feedstock orders three months before delivery, and a series of sourcing errors had eroded its results in the first half of 2020.

How AI helps

Machine learning models are trained on price history and market drivers (exchange prices, technical indicators, economic data, news, supply and demand in user sectors) and give the buyer a forecast of where the price is likely to go over the coming days, weeks or months, often with an explanation of what drives it. The buyer does not hand over the decision: the forecast is one input to decide when to buy and how much of the coming need to cover (buy earlier and more if a rise is expected, wait if a drop is likely), and an argument in supplier negotiations. At La Cibek, Vedrai's virtual agent Becky monitors aluminium prices and suggests purchase times, using historical data, market indicators (Stochastic, MACD, RSI) and macroeconomic news. At Ormazabal, Bork's system gives 10-day-ahead forecasts with response protocols. At the plastics converter, the managing director added ChAI's daily LLDPE forecasts and its AI-generated commentary to her sourcing routine.

Step by step

1. Pick one or two materials that weigh most in your costs and whose price you can follow publicly (an exchange price or an index your suppliers use).
2. Gather 2 to 3 years of your own purchases (date, quantity, price paid) and the market price over the same period; check how closely your supplier prices follow the market price, and with what delay.
3. Write down your buying rules today: how far ahead you must order, how much stock you can hold, how much of the next quarter you usually cover, and who decides.
4. Test a forecasting tool on the past before using it (backtest): would its signals have beaten your actual purchase timing, and how often was it wrong?
5. Run it in parallel for a few months: log each forecast, the decision taken and the price actually paid, and keep a simple rule, for example never cover more than a set share of the need on a single signal.
6. Measure against a fair baseline (for example the average market price over the period, or buying at fixed dates), not against the worst moment, and report the misses as well as the wins.

Tools

• Vedrai (virtual agent Becky for raw-material price monitoring and purchase timing): used at La Cibek
• Bork (machine-learning raw-material price forecasts): piloted at Ormazabal
• ChAI (AI commodity price forecasts with explanations): used at the Asian plastics converter
• Other tools to evaluate: commodity price forecasting services from market-data providers (for example Mintec/Expana), forecasting in your ERP or planning tool, the price indexes your suppliers quote against
• To test the idea without new software: your purchase history and a public price series (for example the LME aluminium price) in a spreadsheet, plus any LLM (ChatGPT, Claude, Copilot, Gemini) with the prompt below

In the AI tools index: Bork · ChAI · Vedrai (Becky) · all AI tools used by manufacturers

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

• La Cibek's figures come from its own news post and cover two purchases only; the plastics converter is not named and its figures are the vendor's; Ormazabal's pilot has no published result.
• A saving measured against what the price did afterwards is hindsight: compare with a fair baseline (average price, buying at fixed dates) over many decisions, and count the times the forecast was wrong.
• Buying early ties up cash and storage, and a forecast can be wrong in either direction; set a limit on how much of the need you cover on one signal, and agree it with finance.
• Forward buying is not hedging: if your selling prices are fixed, talk to your bank or supplier about index-linked prices or hedging before relying on timing alone.

You are a raw-material buyer at a manufacturer.

I will paste: (1) our purchases of [material] over the last 24 to 36 months [date, quantity, unit price paid, supplier]; (2) the market price or index for the same material over the same period [date, price]; (3) our constraints [order lead time, maximum stock we can hold, share of next quarter's need we usually cover, the selling prices or contracts this material goes into].

Do this:
1. Make a table that puts each purchase next to the market price on the same date and one month earlier, and tell me which spreadsheet formulas to use to compute the gap and the correlation; do not compute them yourself.
2. Describe in plain words how our supplier prices seem to follow the market price (closely, with a delay, with a fixed premium), quoting the rows that show it.
3. List our five most expensive purchase moments compared with the market at the time, and what we could have known then.
4. Propose simple buying rules for the next six months (how much to cover, when to review, a maximum share bought on a single price signal) and how to test them on the past before using them.
5. Write 10 questions to ask a price-forecasting vendor (backtest results, forecast horizon, how often it was wrong, explanation of drivers, materials covered, price).

This is not investment advice and you cannot predict prices: do not give a price forecast yourself. Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.

Related cases

• AI watches news and events around the suppliers and sub-suppliers you have mapped, so buyers act before a disruption stops the plant (Renault Group, Kärcher, Woodward, Nissan)
• AI sorts every purchase line into spend categories so buyers see what they really spend with each supplier (Pentair, Barry-Wehmiller)
• AI agents negotiate with long-tail suppliers that buyers never had time to call (Honeywell, and separately an anonymous German consumer and industrial goods maker)
• An AI agent chases a late PO in Zoho Inventory and updates it after approval (Northline Components)

All 13 Procurement use cases →

Sources

• https://lacibekpressofusione.it/en/blog/news-and-events/vedrai-s-artificial-intelligence/
• https://eulacdigitalaccelerator.com/2025/09/15/ormazabal-x-bork-using-ai-to-beat-the-volatility-of-raw-material-prices/
• https://chaipredict.com/resources/securing-better-outcomes-for-plastics-sourcing