How AI Is Changing Commodity Intelligence for Procurement Teams

The old way: expensive data, manual analysis

For decades, procurement teams that wanted commodity intelligence had two options: subscribe to an enterprise data feed, or do without.

Services like Fastmarkets, CRU, and Platts provide excellent data, but they come at a cost. Annual subscriptions typically run £15,000 to £50,000 depending on the metals covered and the depth of analysis included. And even with the data in hand, you still need a skilled analyst to interpret it, map it to your specific parts, and translate it into actionable negotiation positions.

For large enterprises with dedicated commodity teams, this model works. For the thousands of mid-market manufacturers spending £1M to £20M on metal parts annually, it has always been out of reach. These companies buy metal-intensive parts every day but lack the tooling to verify whether the prices they pay are fair.

The new way: AI agents that do the analysis

The shift happening now is not just about cheaper data access. It is about automating the entire analytical workflow that sits between raw commodity data and a procurement decision.

Modern AI agents can:

• Connect directly to market and COMEX APIs for real-time pricing.
• Parse a part number or description to identify the specific metal, alloy, and manufacturing process.
• Build a complete should-cost model from first principles.
• Compare the model to a supplier's quoted price and generate a verdict.
• Draft a counter-offer with specific price references and evidence.

This is not a dashboard you stare at. It is an agent that does the work a commodity analyst would do, in seconds rather than hours.

How Agent Midas works

SupplyVerse's Agent Midas is purpose-built for this workflow. When a buyer submits a part for analysis, Midas follows a structured process:

Step 1: Parse the part

Midas reads the part number, description, material callout, and any referenced standards (NAS, AMS, ASTM, SAE). From this, it identifies the base metal, the specific alloy grade, the likely manufacturing process, and the typical weight range.

Step 2: Identify the metal and pull market pricing

Once the metal is identified, Midas pulls the current market 3-month price for the base metal, then applies alloy surcharges based on the specific grade. For a 7075-T6 aluminium aerospace part, this means starting with market aluminium and adding the premium for the high-zinc alloy content.

Step 3: Build the should-cost model

Midas assembles a full cost stack: material (including scrap allowance), conversion (based on process type and regional rates), freight (based on typical origin-destination pairs), applicable tariffs, and an industry-appropriate margin band.

Step 4: Generate the verdict

The should-cost model is compared to the supplier's quoted price. Midas delivers a clear verdict: fairly priced, slightly over, significantly over, or under-priced (which can indicate quality risk). Every figure is sourced and traceable.

The three-tier confidence system

Not all analyses are created equal. Midas uses a three-tier confidence system to be transparent about the quality of each estimate:

• Verified: The metal grade, alloy, and weight are confirmed from the part specification or drawing. market pricing is live. Confidence is high.
• Estimated: The metal and alloy are inferred from the part number or description, but not confirmed from a specification. The model uses typical weight ranges for the part type. Confidence is moderate.
• Unverified: The part description is ambiguous or the metal cannot be reliably identified. Midas flags this clearly and explains what additional information would improve the analysis.

This transparency is critical. A should-cost model is only useful if the buyer knows how much to trust it.

Bill of Materials analysis at scale

Single-part analysis is useful, but the real power emerges with full BOM processing. Buyers can upload a CSV or Excel file containing their entire bill of materials, and Midas benchmarks every line in parallel.

For a 100-line BOM that would take a human analyst two weeks to should-cost manually, Midas delivers results in under a minute. Every line gets a verdict, a should-cost breakdown, and a variance flag. The buyer can immediately see which lines are fairly priced and which need attention.

This changes the economics of should-cost analysis completely. Instead of focusing on the top 10 parts by spend, teams can verify every single line item.

The challenge email

Perhaps the most practical feature is automated counter-offer drafting. When Midas identifies an overpriced part, it drafts a professional challenge email that:

• References the specific market price and date.
• Breaks down the should-cost components.
• States a specific counter-price with supporting evidence.
• Maintains a professional, collaborative tone.

The buyer can edit the email, adjust the tone, and send it directly. The supplier receives a data-backed counter-offer rather than a vague request for a discount, which leads to faster, more productive negotiations.

Why this matters for SMBs

The companies that benefit most from accessible commodity intelligence are not the large enterprises who already have commodity teams and Fastmarkets subscriptions. They are the mid-market manufacturers who have been priced out of proper tooling until now.

A 200-person aerospace subcontractor spending £5M annually on metal parts cannot justify a £40,000 data subscription and a full-time commodity analyst. But they can justify a tool that gives their existing buyers the same analytical capability at a fraction of the cost.

This democratisation of commodity intelligence is the real shift. The data was always available. The analysis was always possible. What was missing was a tool that made it accessible, fast, and affordable for teams that buy metal parts every day but have never had the means to verify the prices they pay.