Domain I · Area I-CA2 · Responsible AI use, grounding and model-risk control

I.6Ground, verify and document AI-assisted outputs

Working Draft version 1.1

Purpose and scope

This unit covers the capability to test AI-assisted outputs against authoritative sources, underlying evidence, calculations, approved criteria and organisational context. It includes factual, citation, completeness, classification, logic, calculation, consistency, claim, bias and boundary checks; re-performance; contradiction handling; correction; confidence and limitation documentation; and controlled approval and reuse.

Applied competency statement

Can ground and verify AI-assisted reporting outputs against authoritative sources, evidence, criteria and calculations, document corrections and limitations and prevent unverified output from entering approved reporting.

Boundary and escalation

This unit owns AI-specific grounding and verification of source support, completeness, classifications, calculations and context. F.3 owns general reporting-evidence quality and sufficiency, E.7 owns data validation and reconciliation and G.7 owns claim substantiation. It does not provide formal model validation, legal verification, software assurance or independent assurance.

Key quality risks

Fluent output treated as accurate; citations checked only for existence rather than support; one supporting source used while contradictory evidence is ignored; summaries verified by another model rather than the source; calculation and table output not reperformed; completeness and omitted exceptions not tested; model self-confidence treated as evidence; context or entity facts invented; earlier verified output reused after source or model change; reviewer unable to reconstruct prompts and inputs; and unverified output entering a report, decision paper or system.

Required knowledge · 6

I.6-K01

Understands grounding to source, evidence, criteria, calculation and organisational context and the distinction between retrieval, citation and actual support.

Knowledge Type: Grounding concepts · Normative Weight: Core

I.6-K02

Understands factual accuracy, citation support, completeness, classification, logic, arithmetic, consistency, boundary, period, claim, uncertainty and bias checks.

Knowledge Type: Verification dimensions · Normative Weight: Core

I.6-K03

Understands source opening, calculation reperformance, sample and full-population review, independent corroboration, contradiction and alternative explanation.

Knowledge Type: Re-performance and corroboration · Normative Weight: Core

I.6-K04

Understands hallucination, fabricated citation, context loss, omission, prompt sensitivity, model and version variability, stochastic output and automation bias.

Knowledge Type: Output and model limitations · Normative Weight: Core

I.6-K05

Understands correction, re-prompting, manual revision, rejection, approval, expiration, source or model change and the need to reverify reused outputs.

Knowledge Type: Correction and reuse · Normative Weight: Core

I.6-K06

Understands prompt, input, source, model, version, output, verification procedure, finding, correction, reviewer, decision and final-use records and specialist escalation.

Knowledge Type: Documentation and boundary · Normative Weight: Core

Applied skills · 5

I.6-S01

Define material verification assertions, authoritative sources, calculations, sampling, thresholds, reviewer and completion criteria for an AI output.

Skill Type: Core applied capability · Observable Output or Result: AI output grounding and verification plan

I.6-S02

Trace material statements, classifications and citations to authoritative sources and evidence and identify unsupported or contradicted content.

Skill Type: Source and evidence grounding · Observable Output or Result: Source, evidence and calculation grounding matrix

I.6-S03

Reperform material calculations, tables, comparisons and reasoning steps and test completeness, boundary, period and consistency.

Skill Type: Calculation and logic verification · Observable Output or Result: AI calculation and logic verification record

I.6-S04

Correct, qualify or reject unsupported output and document remaining uncertainty, restricted use and conditions for reuse.

Skill Type: Correction and limitation · Observable Output or Result: AI output correction, limitation and approval record

I.6-S05

Obtain substantive human review and approval and reverify when sources, prompts, tools, model versions or use contexts change.

Skill Type: Approval and monitoring · Observable Output or Result: AI output approval and reverification log

Professional behaviours · 3

I.6-B01

Does not transfer verification responsibility to the AI system, another unverified model output or a superficial citation check.

Behaviour Type: Verification ownership · Non-compensable Requirement: No

I.6-B02

Does not retain a convenient AI answer when authoritative or entity evidence contradicts it.

Behaviour Type: Contradictory-evidence integrity · Non-compensable Requirement: No

I.6-B03

Does not allow material AI-assisted content, classification or calculation into approved reporting before required human verification and approval.

Behaviour Type: Release discipline · Non-compensable Requirement: No

Typical tasks · 4

I.6-T01

Identify the AI output, material assertions, intended use, sources, criteria, calculations, risks and required verification depth.

Primary Output Link: I.6-O01

I.6-T02

Open and compare sources, trace citations and evidence and investigate omissions, contradictions and entity-specific context.

Primary Output Link: I.6-O02

I.6-T03

Reperform calculations and logic, test classification, boundaries, periods, claims and consistency and record findings.

Primary Output Link: I.6-O02

I.6-T04

Correct, qualify or reject the output, obtain human approval and control reuse and reverification after change.

Primary Output Link: I.6-O03

Expected outputs · 3

I.6-O01

AI output grounding and verification plan

Output Type: Professional work product

I.6-O02

Source, evidence and calculation grounding matrix

Output Type: Professional work product

I.6-O03

AI output correction, limitation and approval record

Output Type: Professional work product

Proficiency indicators

Level 1 · Foundation

I.6-L1-01

Can apply an approved AI verification checklist, trace specified statements and citations and identify obvious unsupported or inconsistent output.

Indicator Dimension: Task execution

I.6-L1-02

Can stop unverified use and escalate material calculation, legal, technical or contradictory-evidence questions.

Indicator Dimension: Quality, judgement and accountability

Level 2 · Practitioner

I.6-L2-01

Can independently ground and verify a moderately complex AI-assisted work product containing research, classifications, calculations and draft narrative.

Indicator Dimension: Task execution

I.6-L2-02

Can determine verification depth, resolve routine findings and explain corrections, limitations and reuse conditions to management and reviewers.

Indicator Dimension: Quality, judgement and accountability

Level 3 · Advanced Practitioner

I.6-L3-01

Can design or critically review enterprise AI grounding, verification, approval and reverification methods across complex reporting use cases.

Indicator Dimension: Method design and review

I.6-L3-02

Can challenge superficial verification or model-to-model checking, resolve significant contradictory output and advise governance bodies on rejection, restricted use and residual risk.

Indicator Dimension: Leadership and governance

Illustrative evidence · 6

I.6-E01

AI output grounding and verification plan.

Evidence Type: Work product

I.6-E02

Source, evidence and calculation grounding matrix.

Evidence Type: Work product

I.6-E03

AI output correction, limitation and approval record.

Evidence Type: Work product

I.6-E04

Prompt, input, source, citation, reperformance, finding, correction, reviewer, approval and reuse trail.

Evidence Type: Process evidence

I.6-E05

Documented technical, legal, specialist, quality or management review and the practitioner's response.

Evidence Type: Review evidence

I.6-E06

Observed grounding, correction or rejection of a materially unsupported AI output.

Evidence Type: Observed performance

Assessment · 3

I.6-A-L1

AI-output verification exercise, citation tracing and situational judgement

Source and citation support; completeness and calculation; stop-use and escalation.

I.6-A-L2

Integrated AI-verification case and professional memorandum

Verification scope; source and calculation reperformance; contradiction and limitation treatment; approval documentation.

I.6-A-L3

Complex verification-governance case, portfolio and oral defence

Method design; independent grounding and reperformance; contradiction and release decisions; governance advice and oral defence.

Relationships · 8

From	To	Type	Rationale
E.7	I.6	Boundary distinction	E.7 validates and reconciles sustainability data and calculations; I.6 verifies AI-assisted outputs and should invoke E.7 where AI creates, transforms or analyses reporting data.
I.4	I.6	Risk and control linkage	AI-retrieved or summarised information should be grounded and verified before reporting use.
I.5	I.6	Risk and control linkage	AI-assisted mapping, analysis and drafting require source, calculation and context grounding.
I.6	F.3	Evidence and traceability linkage	Grounding and verification should evaluate whether AI-assisted outputs are supported by appropriate reporting evidence.
I.6	E.7	Method, data and analytical linkage	AI-generated classifications, calculations and tables may require data validation, reconciliation and reperformance.
I.6	G.7	Disclosure, claims and publication linkage	AI-assisted claims should not enter reporting unless evidence, scope, qualification and overall impression are verified.
I.6	I.5	Operational and workflow linkage	Verification converts candidate AI-assisted work into a controlled professional work product.
I.6	I.7	Risk and control linkage	Use-case and model risk determine verification assertions, sample, edge cases and reverification triggers.

Role profiles for this unit

Role	Target level	Relevance	Evidence expectation
Corporate Sustainability Reporting Practitioner	Practitioner	Required?	A case or work sample demonstrating independent performance, documented judgement and a reviewable professional output.
Sustainability Reporting Manager or Lead	Advanced Practitioner	Role-defining?	A complex case or verified portfolio, supplemented by oral or observed defence, demonstrating method design, challenge and governance capability.
Sustainability Reporting Adviser or Consultant	Advanced Practitioner	Role-defining?	A complex case or verified portfolio, supplemented by oral or observed defence, demonstrating method design, challenge and governance capability.
Sustainability Data, Systems and Controls Specialist	Advanced Practitioner	Role-defining?	A complex case or verified portfolio, supplemented by oral or observed defence, demonstrating method design, challenge and governance capability.
Assurance Readiness and Reporting Quality Specialist	Advanced Practitioner	Role-defining?	A complex case or verified portfolio, supplemented by oral or observed defence, demonstrating method design, challenge and governance capability.
Investor, Capital Markets and Ratings Disclosure Specialist	Advanced Practitioner	Role-defining?	A complex case or verified portfolio, supplemented by oral or observed defence, demonstrating method design, challenge and governance capability.