How many trips does a reusable crate have to make?

A supplier quotes a reusable crate designed to last 30 trips. A buyer asks an agent to compare it with a single-use carton. At 30 uses the reusable option wins on the assumptions below. At ten, it loses.

The question the agent needs to ask is simple: how many times does the crate actually come back? A claimed service life is only one part of the answer. Damage, missed returns, cleaning and the distance to the return hub can change the calculation.

Procurement teams can use an agent to surface those questions before ranking a bid. For the shortlist to survive a review, the buyer also needs to know where each number came from and which comparisons the data supports.

What corporate disclosure can tell a buyer

BCG and CDP's analysis of 2023 disclosures found that corporate supply-chain emissions were 26 times operational emissions on average in Exhibit 1, which filters to 2,229 companies with sufficient evidence of verification. The authors also reported that 15% of disclosing companies had set a supply-chain emissions target. The first is a corporate reporting ratio, not a multiplier for an individual product. The second is a share of disclosers, with the limitations of a voluntary reporting population. Neither tells a buyer how many comparable footprints a specific supplier market can provide.

The GHG Protocol Product Standard covers life cycle accounting and emphasizes the role of product rules when comparing products. WBCSD's PACT Methodology V3 specifies how to calculate and exchange cradle-to-gate product carbon footprints. This narrower boundary can support data exchange between suppliers and buyers. If a purchasing decision depends on use, repair, reuse or disposal, the buyer also needs compatible evidence for those stages.

These sources suggest a testable near-term opportunity: use agents to raise the share of offers backed by comparable evidence and reduce reviewer time. There is no adoption or emissions reduction estimate for agentic procurement in these sources. Commercial and environmental outcomes require records of what was purchased and how it was used.

Run the numbers on 1,000 deliveries

Suppose a buyer needs 1,000 equivalent deliveries. In this illustration, a single-use container has a modeled footprint of 0.45 kg CO₂e per delivery, or 450 kg CO₂e for the service. A reusable container has 4.8 kg CO₂e of production and end-of-life impact and 0.14 kg CO₂e in cleaning and return transport per delivery. Assume equivalent outbound transport, functionality and loss treatment. On those assumptions, reusable first beats the single-use reference at 16 whole uses, because 4.8 ÷ (0.45 − 0.14) ≈ 15.5.

The calculation allocates manufacturing and end-of-life impact across actual uses and adds per-delivery operations. Values are illustrative and rounded. A supplier may advertise a 30-use design lifetime while field losses and damaged returns yield only ten actual uses. That difference reverses the comparison. If cleaning and return transport rise to 0.25 kg per delivery, the break-even moves to 24 uses for parity. The agent should display these assumptions and preserve the calculation for review.

Put the comparison rules in the request for quotes

• Set functional requirements first. Compare capacity, strength, delivery conditions and safety against the buyer's requirements. Measure cost per completed delivery, including reverse logistics and loss.
• Define comparability. Match product function, unit, boundary, geography, method and reporting period. Log primary data share and verification status where available. A missing value remains unknown.
• Preserve the evidence trail. Store document version, product and supplier IDs, extraction location and review status. Do not convert a corporate Scope 3 total into a SKU footprint.
• Constrain action. Allow the agent to request missing information and prepare a shortlist for approval. Give the buyer and sustainability reviewer access to the sources and assumptions that affect ranking.

What would change the buying decision at scale?

In the near term, the best use of an agent may be finding gaps in supplier evidence and sending precise follow-up questions. If suppliers start exchanging compatible product footprints and buyers collect field data on use and returns, carbon can influence more decisions alongside cost and service.

A useful 2030 indicator is spend covered by comparable product-level evidence, broken down by category. Where that share stays low, fund better measurement for the purchases where the answer could change the award. More uploaded documents will help only when a reviewer can reproduce the comparison.

Give the shortlist to two reviewers

1. Select one category with several eligible suppliers and a clear functional unit. Record the manual baseline for review time, coverage and agreement between reviewers.
2. Build a review set containing duplicate documents, mismatched boundaries, obsolete versions, rounded values and missing disclosures. Have procurement and sustainability reviewers label comparability independently.
3. Run the agent in shadow mode. Measure extraction accuracy, comparable-evidence coverage, false comparisons and reproducibility. Include abstentions in the denominator for all eligible offers.
4. Proceed to assisted purchasing only if the shortlist is reproducible and material evidence gaps reliably trigger human review. Track actual cost, delivered service and realized product use. Report emissions estimates with the source data, assumptions and uncertainty needed to reproduce them.

Evidence limits. BCG–CDP data describe corporate disclosure. GHG Protocol and PACT provide accounting guidance, and the packaging example illustrates sensitivity to real-world usage. It is neither a supplier assessment nor a claim of measured emissions savings.

Sources and related research

• BCG and CDP, Scope 3 Upstream: Big Challenges, Simple Remedies, 2024 — disclosure analysis and target setting.
• GHG Protocol, Product Life Cycle Accounting and Reporting Standard — boundaries and product rules.
• WBCSD, PACT Methodology Version 3 — cradle-to-gate accounting and exchange.
• GAEO.ai, applied AI and agent commerce evidence brief and research across industries.