COMPEL Specialization — AITE-VDT: AI Value & Analytics Expert Template 4 of 5

This template produces a VRR to Article 16’s six-section specification. The output is typically 8–15 pages for a single feature — executive-summary consumers read only Section 1, audit-committee consumers read the full document, regulators consume the document plus appendices.

Value Realization Report: [Feature Name]

Feature: [Short name] Reporting period: [Q1 2026 / Month YYYY] Report version: 1.0 Date of this report: [Date] Prepared by: [AI value lead, name] Reviewed by: [List of reviewers: FinOps lead, compliance, communications] Aligned to: ISO 42001 Clause 9.3; NIST AI RMF MEASURE function

1. Executive summary

Feature: [One sentence: what the feature does, who uses it, what outcome it targets.]

Reporting period realized value: [$ amount — one number, with counterfactual method named in parens]

Status: [Green / Yellow / Red]

Primary finding: [One sentence — the headline of the quarter]

Decision requested (if applicable): [One sentence — the specific decision the reader is being asked to make, with date.]

Portfolio context

• Feature’s position on quarterly portfolio scorecard: [Reference]
• Cumulative realized value to date: [$]
• Cumulative investment to date: [$]
• Current cumulative payback ratio: [X.Y]

What changed since last report

[Two to three sentences on significant changes: launches, incidents, decisions, environment shifts.]

2. KPI tree and realized values

KPI tree reference: [Link or embedded visualization of the tree]

Outcome-level (Level 1): [Name]

Driver-level (Level 2):

Leaf metrics (Level 3): [Table or embedded dashboard link]

Evidence notes

• Reproducibility: [Statement that every metric is reproducible from source data, with any gaps disclosed.]
• Source-system status: [Any source-system issues during the period that affected metric quality.]
• Methodology changes: [Any changes to metric computation this period and their impact.]

3. Counterfactual narrative

Counterfactual method: [A/B test / DiD / RDD / Synthetic control / PSM / Pre/post]

Rationale for method choice: [One paragraph — why this design is appropriate for this feature’s rollout shape and data availability.]

Specification: [Brief statement of the model specification — the equation, the estimator, the clustering.]

Point estimate: [X% or $X]

Uncertainty band: [95% CI: [low] — [high]; or p10/p50/p90 from Monte Carlo]

Robustness checks performed:

• [Parallel-trends / placebo / bandwidth / leave-one-out / Rosenbaum bounds]
• [Result of each check]

Known limitations:

• [Limitation 1 and its disclosure]
• [Limitation 2 and its disclosure]

Narrative interpretation: [One paragraph that translates the quantitative result into business language, with uncertainty preserved. Example: “Under the DiD specification, the feature contributed a 6% improvement in the primary metric, with 95% CI 3.5%–8.5%. At current scale, this translates to approximately $14M in annualized realized value. The estimate is robust to excluding the non-randomized pilot office and to alternative estimators; it is less confident in the lower-than-average practice areas, where separate analyses are ongoing.”]

4. Financial summary

Realized value

Attribution-sensitivity: [Same figures under 1 or 2 alternative attribution models, if material.]

Total cost of ownership (TCO)

rNPV trajectory

• Original business-case rNPV: [$]
• Current estimated rNPV based on realized-data update: [$]
• Variance: [$ or %]
• Driver of variance: [Higher/lower benefit than projected / cost overrun / timing shift]

Budget-vs-actual (compute)

• Budget this period: [$]
• Actual this period: [$]
• Variance: [$ or %]
• Enforcement actions during period: [Alert / Throttle / Reject / Review]

Sustainability-Adjusted Value (SAV)

• Core realized value: [$]
• Externality adjustment: [$ — with breakdown if material]
• SAV: [$]
• Trajectory vs. prior period: [+/-%]

5. Risk flags

Value risks

Operational risks

Governance risks

Risks materialized this period

[List any risks that were hypothetical in last report and now are actual.]

Risks resolved this period

[List any risks that have been closed.]

6. Recommendation

Recommendation: [Continue / change / stop]

Detail:

• [Continue:] [One paragraph — why, with reference to the analysis]
• [Change:] [One paragraph — what scope, design, or operational change is proposed and why]
• [Stop:] [One paragraph — sunset case reference; see Template 5 equivalent for sunset case structure]

Supporting evidence: [Reference sections 2, 3, 4, 5 of this report]

Dissent (if any): [Any reviewer whose disagreement should be recorded in the decision log]

Decision log entry

Appendix A — Detailed tables

[Full metric table with all leaf metrics, all periods, and full computation trail. Multi-page if needed.]

Appendix B — Analysis methodology detail

[Full mathematical specification of the counterfactual analysis, diagnostic-test results, robustness-check outputs, sensitivity analyses.]

Appendix C — Source documents

• Measurement plan: [Reference]
• Business case (original): [Reference]
• Business case (most recent update): [Reference]
• KPI tree: [Reference]
• Compute budget: [Reference]
• Prior VRRs: [References]

Appendix D — Regulatory and standards alignment