CMSI AI Assessor Pre-auditAI pre-audit for the Consolidated Mining Standard
Upload your facility's policies, plans and reports — in any language. CMSI AI Assessor evaluates all 48 sub-sections across the 24 Performance Areas of the Consolidated Standard, producing a formal pre-audit with evidence, gaps and recommendations mapped to the three CMSI performance levels.
From documents to a formal pre-audit
A three-step pipeline that mirrors the CMSI Assurance Process.
Upload to the document library
Drop in any number of documents — PDF, DOCX, XLSX, PPTX (up to 50 MB each). Each upload is auto-summarized, tagged with key topics, and classified against the 48 sub-sections so the system knows which sub-sections it carries substantive evidence for.
Run an assessment
Name the facility, optionally flag Not Applicable sub-sections, pick the documents, and click Begin Assessment. The system evaluates all applicable sub-sections in batches of 6 concurrent runs, streaming findings to the dashboard live.
Read the formal report
An editorial-style printable report with cover page, outcome banner, Claims Policy criteria checklist, executive summary, PA-by-PA deep dive with per-sub-section findings. Export as PDF using your browser's Print to PDF.
Four Pillars. Twenty-four Performance Areas. Forty-eight sub-sections.
The Consolidated Mining Standard organises responsible mining performance across four Pillars, each of which contains six PAs covering distinct ESG topics.
Ethical Business Conduct
Corporate accountability, business integrity and anti-corruption, transparent payments, responsible supply chain, due diligence in conflict-affected and high-risk areas.
Worker & Social Engagement
Labour rights, occupational health and safety, security and human rights, stakeholder engagement, grievance mechanisms, indigenous peoples' rights and free prior informed consent.
Social Performance
Community development and benefits, resettlement and land acquisition, cultural heritage, artisanal and small-scale mining, mine closure planning, decommissioning and post-closure stewardship.
Environmental Responsibility
Tailings management, water stewardship, biodiversity and ecosystems, climate change, hazardous materials, air and noise emissions, smelter and refining requirements.
CMSI's three-level performance ladder
Each sub-section receives one of six states. The level is derived deterministically from per-requirement determinations — the LLM evaluates evidence per requirement, code derives the level. No subjective level guessing.
All TGP + GP + LP requirements demonstrated
The facility evidences requirements at all three levels — full mastery of the sub-section.
All TGP + GP demonstrated
One or more Leading Practice requirements not yet met. Solid conformance.
All TGP requirements demonstrated
One or more Good Practice requirements not yet met. The minimum acceptable level.
One or more TGP requirements unmet
The only "warning" register. Renders in Oxblood. Blocks Performance Claim Eligibility.
No documents address this sub-section
Surfaced as a soft flag. Does not count toward a formal level. Lists expected-but-absent requirements.
Flagged by the facility
Not relevant to operations (e.g., PA 23.2 Smelter requirements at a non-smelter site). Rationale recorded.
Per-requirement determinations drive the level
For each Towards Good Practice / Good Practice / Leading Practice requirement defined in the standard, the LLM marks Met, Partially Met, Not Met or Not Addressed — with a one-line rationale. Code then derives the achieved level deterministically. This is the level of detail an Assurance Provider expects.
The Claims Policy outcome
CMSI's Claims Policy specifies the conditions under which a facility may make a Performance Claim. CMSI AI Assessor reports which conditions pass and which block the claim, so the compliance team knows exactly what is needed to advance.
Aggregate score ≥ 80%
Sum of sub-section scores divided by maximum possible (with Leading Practice = 3 points, Good Practice = 2, Towards Good Practice = 1, Below TGP = 0). Insufficient Evidence and Not Applicable sub-sections excluded from the denominator.
Every applicable PA at TGP or higher
Each PA's score is the lowest level achieved across its applicable sub-sections. A single Below-TGP sub-section blocks its entire PA.
≥ 80% of PAs at Good Practice or higher
Of the applicable PAs, at least 80% must roll up to Good Practice or above. This rewards consistent above-minimum performance.
≥ 4 PAs per Pillar at Good Practice or higher
Each of the four Pillars must have at least four PAs at Good Practice. Prevents a facility from compensating weakness in one Pillar with excellence in another.
Below TGP → 0 | TGP → 1 | Good Practice → 2 | Leading Practice → 3
The four meta-outcomes
Once all sub-sections resolve, the system computes one of four overall determinations.
Performance Claim Eligible
All four Claims Policy criteria pass. The facility may advance to external assurance to support a Performance Claim.
Continual Improvement Required
All applicable PAs are at TGP or higher, but one or more Claims Policy thresholds are not met. The dashboard shows exactly which.
Below Standard
One or more PAs are Below TGP. The facility is not yet ready for external assurance and must address the unmet TGP requirements.
Indeterminate
Sufficient sub-sections are at Insufficient Evidence (> 20% of applicable) that a meaningful overall outcome cannot be calculated. Upload more documents.
Every sub-section produces five structured outputs
No paraphrased fluff. Direct quotes from your documents, attributed to the source, with verifiable gaps cited to specific requirements and one-to-one recommendations.
Level + auditor-voice rationale
One of the six states above, plus a 2–3 sentence rationale in the auditor voice explaining why this level was selected for this sub-section.
Per-requirement marks
Every TGP, GP and LP requirement in the standard marked Met / Partially Met / Not Met / Not Addressed with a one-line rationale. The level of detail an Assurance Provider expects.
Verbatim quotes, attributed
Direct verbatim quotes from your submitted documents that support the verdict. Each is attributed to the source document and the requirement it evidences. No paraphrase.
Specific missing requirements
Actionable items from the standard's requirements, missing or underdeveloped in the documentation. Cited to specific requirement IDs (e.g., "GP requirement 1.4.3"), not generic observations.
Concrete actions, mapped to gaps
One-to-one mapping with the identified gaps wherever possible. Concrete, scoped, owner-friendly actions you can take into the next assurance cycle.
A live assessment dashboard and a printable formal report
Watch the assessment run sub-section by sub-section. Then read the editorial-style report with Foolscap-on-Slate typography.
Assessment dashboard
Outcome banner at the top with the Foolscap banner styling — restrained Oxblood seal for Performance Claim Eligible; Slate-bordered banners for the other states.
Claims Policy checklist showing which of the four conditions passed or blocked the claim.
Pillar / PA / Sub-section rail with all 48 sub-sections grouped by Pillar → PA → sub-section, each with a verdict bar.
Detail panel showing the selected sub-section's full findings.
Coverage panel with documents-by-sub-section matrix, evidence quote thumbnails, and a small radar chart across the four Pillars.
Formal report
Editorial typography with Foolscap-on-Slate page styling for a printable, high-craft feel.
Cover page with the outcome banner and overall determination.
Executive summary with the Claims Policy criteria checklist for leadership and assurance teams.
PA-by-PA deep dive with per-sub-section findings (achieved level, per-requirement determinations, evidence quotes, gaps, recommendations).
Print to PDF via the browser. Server-side PDF generation is on the roadmap.
Pre-loaded with a realistic demo case
When you open the app, it is pre-populated with a fictitious mining facility — including a corporate document library, a completed assessment with all 48 sub-sections resolved, and a realistic mix of levels across the four Pillars.
Facility
A fictitious mining facility used to showcase real-world structure: governance, social and environmental scope plus an optional smelter site.
Corporate document library
Code of conduct, anti-bribery, human rights, community engagement, indigenous peoples, tailings management, water stewardship, biodiversity, climate, mine closure, supplier code, OHS and grievance mechanism.
Pre-computed assessment
All 48 sub-sections resolved across the four Pillars. A balanced mix of Leading Practice highlights, Good Practice mainstream and a handful of Towards Good Practice gaps to address.
Inspect the existing assessment, then upload your own documents and run a fresh assessment against the live LLM pipeline.
Common questions
Is CMSI AI Assessor affiliated with the Consolidated Mining Standard Initiative?
No. CMSI AI Assessor is an independent pre-audit instrument built by Data Riders. It is not affiliated with the Consolidated Mining Standard Initiative, ICMM, World Gold Council, The Copper Mark, or the Mining Association of Canada. Reports include a clear disclaimer.
What does the assessment cover?
All 48 sub-sections across the 24 Performance Areas of the CMSI Consolidated Standard, distributed across four Pillars: Ethical Business Conduct (PAs 1–5), Worker & Social Engagement (PAs 6–11), Social Performance (PAs 12–17), and Environmental Responsibility (PAs 18–24). Each sub-section is evaluated using the official three-level performance ladder.
What languages are supported?
Documents can be submitted in any language. Findings are written in the language of the submitted documents (with an optional user override). The interface ships in English; full UI internationalisation is on the roadmap.
How are levels determined?
The LLM evaluates each individual requirement (TGP, GP, LP) as Met, Partially Met, Not Met or Not Addressed. The achieved level is then derived deterministically by code: all TGP met → at least Towards Good Practice; all TGP + GP met → Good Practice; all three levels met → Leading Practice; any TGP not met → Below TGP. The LLM evaluates evidence per requirement; level derivation is code, not LLM.
What is the Claims Policy outcome?
Once all applicable sub-sections resolve, the system computes the overall outcome against four CMSI Claims Policy thresholds: aggregate score ≥ 80%, every applicable PA at TGP or higher, ≥ 80% of PAs at Good Practice or higher, and at least 4 PAs per Pillar at Good Practice or higher. The four meta-outcomes are Performance Claim Eligible, Continual Improvement Required, Below Standard, or Indeterminate.
How does Not Applicable work?
The user can flag sub-sections as Not Applicable when they are not relevant to the operation (e.g., PA 23.2 Smelter requirements at a non-smelter site). NA sub-sections are excluded from numerator and denominator of the aggregate score. The rationale is recorded for transparency.
Do I need to log in?
No. CMSI AI Assessor runs in demo-mode by design: no authentication, single shared workspace. Documents and assessments persist across sessions.
Does the AI replace a real CMSI assurance engagement?
No. CMSI AI Assessor is a pre-audit tool that identifies gaps and surfaces evidence before a formal assessment by an Assurance Provider. The final report includes a disclaimer that the AI assessment is for pre-audit use only and does not constitute a CMSI external assurance engagement.
What file types are supported?
PDF, DOCX, XLSX and PPTX, up to 50 MB each. On upload, full text is extracted, a 1–2 sentence summary is generated, key topics are tagged, and the document is classified against the 48 sub-sections so the system knows which sub-sections it carries substantive evidence for.
Other agents in the compliance family
Related reads & cases
Deep dives on the Consolidated Mining Standard and how AI pre-audits work in practice.
Run your first CMSI pre-audit today
No procurement. No login. Open the assessor, inspect the demo case, then upload your own documents and run a fresh evaluation against the live pipeline.
CMSI AI Assessor is an independent pre-audit instrument built by Data Riders. Not affiliated with the Consolidated Mining Standard Initiative, ICMM, World Gold Council, The Copper Mark, or the Mining Association of Canada.