ISAE 3000 walkthrough¶

A walkthrough of the assurance review your sustainability assurance partner will conduct against your evidence pack.

What ISAE 3000 limited assurance is¶

ISAE 3000 (Revised) is the international standard for assurance engagements other than audits or reviews of historical financial information. Limited assurance means the assurance partner provides a negative opinion: "based on the work performed, nothing has come to our attention that causes us to believe that [X] has not been prepared, in all material respects, in accordance with the applicable criteria."

This is the level of assurance that fits the maturity of AI scope-3 disclosure today. It is also the level most CSRD reporters are working at, before transitioning to reasonable assurance later.

What the assurance partner reviews¶

A typical engagement walks through:

Stage 1 — Methodology assessment¶

Read the methodology document at the version cited
Verify version-pinning (the document version cited matches the bundle's contents)
Assess methodology against ISO 14040/14044 framework
Confirm AGPL-3.0 license (open-source methodology is increasingly an assurance expectation)
Identify any methodology deviations from prior period; assess whether deviations are appropriately disclosed

Stage 2 — Calculation lineage¶

Sample receipts from the period (typical sample 30–100 receipts, stratified by region and model)
Re-compute the sampled receipts from inputs using the methodology code
Verify match within calibration tolerance
Trace each emission factor to its cited source (ecoinvent v3.10, ADEME 2024, etc.)
Verify version pinning of emission factor sources

Stage 3 — Boundary and assumptions¶

Review the boundary statement (narrow vs. comprehensive)
Review the assumption log
Test sensitivity to material assumptions (PUE, embodied lifetime)
Confirm sub-processor stability (no boundary-affecting changes mid-period)

Stage 4 — Aggregation¶

Re-aggregate per-receipt records into the period aggregates
Verify the dual-method scope-2 split (location vs. market-based)
Verify water and resource indicators
Confirm aggregate uncertainty propagation

Stage 5 — Disclosure language¶

Review the customer's narrative against the methodology
Flag any narrative claims that exceed methodology support
Suggest narrative changes where appropriate

Stage 6 — Opinion¶

The assurance partner issues a limited assurance opinion at the entity level, with our methodology as the basis. The opinion typically reads:

Based on the procedures performed and evidence obtained, nothing has come to our attention that causes us to believe that the AI-inference scope-3 disclosures of [Customer] for the year ended [date], have not been prepared, in all material respects, in accordance with the Vetted Inference v[X.Y.Z] methodology and the applicable ESRS criteria.

What we provide to support the assurance¶

Methodology document at all versions in scope of the period (embedded in evidence pack)
Per-receipt records with full calculation lineage
Calibration runs that produced conformal intervals for the period
Pedigree-score worksheet
Version-pinned emission factor sources
Sub-processor change log
Assumption log
On request: direct call between Vetted Inference methodology lead and the assurance partner's review team
On request: site visit / methodology demo (typically remote)

What the assurance partner expects from the customer¶

Reporting boundary clearly stated (which subsidiaries, which use cases)
Disclosure narrative consistent with methodology
Internal controls over the input data (which queries were attributed to which reporting boundary)
Material assumptions documented

Common findings and how to address them¶

Finding: "Methodology is open-source but customer cannot demonstrate they reviewed it"¶

Address: include methodology PDF in your reporting binder; document a methodology-review meeting with your sustainability team.

Finding: "Boundary is comprehensive but disclosure narrative quotes accelerator-only number"¶

Address: revise narrative to cite the comprehensive number; if the narrow number is needed for vendor comparisons, footnote both.

Finding: "Per-receipt uncertainty is wide; aggregate uncertainty is narrow; relationship not explained"¶

Address: add a footnote explaining that independent per-receipt uncertainty propagates to aggregate uncertainty narrowing as 1/√N; the methodology document section 4.3 covers this.

Finding: "Sub-processor change occurred mid-period; aggregate uses pre-change emission factors"¶

Address: re-aggregate using period-appropriate factors (we provide this on request); footnote the change.

Cadence¶

Quarterly: evidence pack delivered (no opinion attached)
Annual: ISAE 3000 limited assurance opinion attached to the year-end aggregate evidence pack
First formal opinion: targeted for FY2027 close (Q1 2028 publication)

Where this is implemented¶

The supporting code is at methodology/aggregation/. The narrative templates are at methodology/templates/isae_3000/.