01 step
Upload or paste
Excel, CSV, or pasted text. No schema required. Cortex reads columns as a colleague would — names, specs, quantities, whatever free-form notes came with the BOM.
Solutions · BOM Matching
Every line in your BOM deserves the same question an expert would ask.
Cortex reads a Bill of Materials the way a practiced LCA consultant reads it — noticing which rows are clear, which are underspecified, and which need to be argued about before any number is assigned.
The flow
§ I
Five steps, always in order.
02 step
Clarify where it matters
For ambiguous rows — "aluminum", "steel", "plastic" — Cortex asks one or two questions before searching. Alloy grade, production route, region. Silence on low-signal fields.
03 step
Search wide, score honest
HiQLCD, Ecoinvent, EF, CarbonMinds, and more — searched in parallel. Each candidate comes back with a Data Quality Indicator (DQI), system model, and a clickable source URL.
04 step
Human-in-the-loop at the thresholds
Three hard gates pause automation: coverage below 80%, proxy substitution, or cross-database GWP spread over 2×. You decide — the agent doesn't silently pick.
05 step
Deliver with provenance
Export to Excel with matched dataset, source database, region, DQI, system model, match type (exact / proxy / low-DQI), and notes. Every cell auditable back to the original factor.
How we differ
§ II
Speed without provenance is a number you can't defend.
Most AI-for-PCF tools race to a single number. We race to a line of reasoning you can hand a regulator. Here's where the two approaches diverge.
Typical AI-for-PCF
Upload → minutes → one PCF number.
Cortex § 1
Upload → clarifying questions → top-k candidates → your approved match → PCF with every assumption attached.
noteSpeed is a side effect. Auditability is the product.
Typical AI-for-PCF
AI decomposes the product automatically.
Cortex § 2
AI proposes. The expert decides. Coverage, proxy use, and cross-database disagreements are never silent.
noteIf a machine can't explain why it picked factor A over B, neither can you when the auditor asks.
Typical AI-for-PCF
Confidence score from a model.
Cortex § 3
DQI — the ISO-recognized Data Quality Indicator. Same vocabulary your reviewers use.
noteUse the industry's words. Don't invent new ones that need translating.
The output
§ III
One spreadsheet. Ten columns of accountability.
Every export contains the fields reviewers look for first: matched dataset, source database, region, DQI, system model, match type, proxy notes. The coverage gap is its own sheet, not a footnote.
Exported fields
- 01 Material name (as provided) col.
- 02 Matched dataset key + URL col.
- 03 Source database (HiQLCD / Ecoinvent / EF / CarbonMinds) col.
- 04 Region · Unit · System model col.
- 05 DQI score col.
- 06 Match type: exact · proxy · low-DQI col.
- 07 Notes (what was substituted and why) col.