I-AVAM™ 2.0: Issifu AI Value Assessment Matrix
A gated, stage-sensitive framework for AI investment and scale decisions.
I-AVAM™ 2.0 helps executives decide whether an AI initiative should move from discovery to experiment, pilot, controlled deployment or scale, and ends in an auditable recommendation, risk statement and remediation plan.
Attractive Ideas Are Not the Same as Investable Value
AI investment decisions rarely fail because organizations lack ideas.
They fail when attractive possibilities move forward before leaders know whether the problem is strategically important, whether the solution can be delivered and sustained, whether the evidence is sufficient for the decision being made, and whether the intended use can be trusted in its operating context.
These questions get harder as initiatives move from experimentation into production. At scale, technical feasibility is no longer enough: scale requires ownership, operational support, assurance controls, benefit measurement and monitoring.
I-AVAM™ 2.0 does not replace enterprise architecture, legal assessment, model validation, cybersecurity review, responsible-AI governance or financial appraisal. It brings their outputs into one investment-stage conversation.
Four Separations That Make AI Decisions Clearer
Comparison Is Not Permission
A score ranks eligible initiatives. Mandatory gates decide whether an initiative may progress at all, so a strong commercial estimate cannot mask a material regulatory, harm, data-rights, ownership, evidence or assurance weakness.
Evidence Is Judged Against the Stage
Discovery does not need the evidence that scale needs. Every stage needs evidence appropriate to the commitment being requested.
Readiness Is Not Portfolio Funding
An initiative can be ready yet still be deferred because people, platforms, budget or dependencies are committed elsewhere. That is an allocation choice, not a readiness failure.
Expected Value Is Not Realized Value
Forecast credibility is assessed before deployment. Attributable benefits are measured after it, and the comparison improves future forecasts.
Four Questions Every AI Investment Must Answer
Every indicator has one primary home, so the same weakness is not counted twice under different labels.
Strategic Attractiveness
Is the opportunity worth pursuing?
Strategic alignment, problem significance, expected benefit, adoption value, time to value and affordability. It records the value proposition, kept separate from the strength of evidence behind it.
Delivery Readiness
Can the initiative be delivered and sustained?
Data, technical and process readiness, execution capability, accountable ownership and change readiness for this specific initiative.
Evidence Confidence
Is the information adequate for this stage?
Problem evidence, feasibility tests, prototype or pilot results, support for benefit estimates, transparent assumptions, replication and independent challenge.
Trustworthiness
Are risk and control conditions acceptable?
Reliability, safety, fairness, privacy, security, transparency, human oversight, accountability, monitoring and incident response. Trust is shown through documented assurance evidence, not asserted by the initiative team alone.
Nine Layers, Nine Distinct Decisions
Each layer answers a different question and produces a different output, so concerns are never collapsed into a single number.
| # | Layer | Decision question | Output |
|---|---|---|---|
| 0 | Regulatory & harm screen | Is the intended use permissible, and what level of assurance is required? | Risk classification, prohibited-use check, specialist-review requirement |
| 1 | Strategic Attractiveness | Is the opportunity worth pursuing? | Strategic and economic rationale |
| 2 | Delivery Readiness | Can the initiative be implemented and sustained? | Initiative-level execution capability |
| 3 | Evidence Confidence | Is the supporting information sufficient for the current stage? | Stage-relative evidence ratio and confidence |
| 4 | Trustworthiness | Are the risks and controls acceptable in the intended context? | Assurance profile and residual-risk position |
| 5 | Mandatory gates | Do non-negotiable progression conditions pass? | Pass, conditional pass, remediation hold or stop |
| 6 | Scale Decision Score | How does a gate-qualified initiative compare with alternatives? | Transparent comparison score |
| 7 | Portfolio selection | Should the organization fund or sequence the initiative now? | Funding, sequencing or deferral decision |
| 8 | Post-deployment review | Did it create attributable value and stay within control expectations? | Benefit, adoption, incident and sustainability learning |
Six Conditions a High Score Cannot Override
A weighted score allows comparison among initiatives. Mandatory gates stop unacceptable weaknesses from being averaged away: a strong value proposition does not offset unlawful use, unmitigable harm or the absence of accountable ownership.
A Conditional Pass requires named actions, owners, deadlines, accepted residual risk and a reassessment point. Passing every gate establishes eligibility for the next decision. It does not commit investment.
- Regulatory permissibilityThe intended use is lawful and all applicable classification obligations are identified.
- Critical harmNo unmitigable material harm condition prevents progression.
- Data rightsLawful access, permitted use, provenance and retention arrangements are in place.
- Accountable ownershipNamed business, technical, risk and benefit owners accept their responsibilities.
- Stage evidence minimumRequired evidence for the current stage is available and reviewable.
- Trust minimumAssurance controls meet the minimum requirement for the intended use and stage.
The Right Evidence for the Commitment Requested
The question is never whether an initiative is universally ready, but whether it has enough value, readiness, evidence and assurance for the next stage.
Discovery
Problem definition, stakeholder need, strategic relevance, initial risk classification.
Fund investigation
Experiment
Prototype results, data and technical feasibility, initial assurance review.
Fund a controlled pilot
Pilot
Operational results, user response, benefit indicators, control testing.
Authorize controlled deployment
Controlled Deployment
Repeated performance, integration evidence, operating ownership, monitoring design.
Consider broader scale
Scale
Stable outcomes, approved controls, support model, validated benefit case.
Authorize wider rollout
Post-Deployment
Attributable benefits, adoption, incidents, drift, operational sustainability.
Continue, expand, modify or retire
Ready Is Not the Same as Funded Now
Delivery Readiness asks whether a particular initiative can move forward. Portfolio selection asks whether the organization can resource it now, alongside every other eligible initiative competing for the same people, platforms, data services and capital.
One Page from Analysis to Governance Action
The canvas turns assessment results into an executive decision record. It adds no new score; its job is to make the decision logic visible, explainable and actionable.
- The four-construct profile, read together with the gate outcome
- Strongest and weakest dimensions and the primary gap
- A primary risk statement and stage-progression recommendation
- Remediation commitments with owners and deadlines
A balanced profile cannot reverse a failed mandatory condition, and a passing gate does not fund anything.
- Define the decisionInitiative, intended use, lifecycle stage, requested commitment and decision authority.
- Screen the useRegulatory classification, critical harms, data rights and specialist reviews.
- Assemble evidenceStage-appropriate business, technical, operational and assurance evidence.
- Score independentlyThe initiative owner proposes ratings; an independent challenger reviews them.
- Apply gatesRecord the outcome for every mandatory condition.
- Compare eligible initiativesScale Decision Score, dimension results, sensitivity and convergence.
- Select at portfolio levelAffordability, resources, dependencies, sequencing, opportunity cost.
- Record and reassessDecision, actions, owners, dates, residual risk and next review.
- Review realized outcomesBenefits, adoption, incidents, drift and sustainability.
Higher-risk initiatives are reviewed by a multidisciplinary group: business ownership, technology, data, security, privacy, legal, responsible AI, finance, operational risk and affected users.
What Leaders Say About I-AVAM™
Experiences from executives and teams who have used I-AVAM™ in their organizations.
Engagement Options
I-AVAM™ Executive Assessment
Applies I-AVAM™ 2.0 to a single initiative, ending in an Executive Canvas with a recommendation, risk statement and remediation plan.
AI Portfolio Value Review
Compares gate-qualified initiatives and supports funding, sequencing and deferral decisions across the portfolio.
AI Pilot-to-Scale Diagnostic
Tests whether stage evidence, delivery readiness and trustworthiness support moving from pilot to controlled deployment or scale.
AI Investment Governance Design
Designs stage gates, decision rights, independent challenge and the minimum decision record.
AI Value Realization Office
Tracks attributable benefits, adoption, incidents and drift after deployment to improve future forecasts.
Executive AI Decision Workshop
Brings leadership together to apply I-AVAM™ 2.0 to live investment and scale decisions.
Open About What Is Still Being Tested
I-AVAM™ 2.0 is a conceptual decision architecture, set out in a journal-ready preprint (October 2026). Its value lies in transparency, consistency, challenge and traceability, not mathematical certainty. Weights, thresholds and predictive performance are being calibrated through a defined research programme:
- Content validity through a multidisciplinary expert panel
- Inter-rater reliability across independent assessors
- Retrospective testing against completed initiatives
- Baseline comparison with value-effort scoring, checklists and expert judgment
- Sensitivity analysis, a prospective pilot and transparent recalibration
I-AVAM™ 2.0 complements the NIST AI Risk Management Framework, ISO/IEC 42001 and the EU AI Act by using the evidence they produce. It is a decision aid, not a substitute for legal advice, technical validation or executive accountability.
What is I-AVAM™ 2.0?
A gated, stage-sensitive framework for AI investment and scale decisions. It assesses four constructs (Strategic Attractiveness, Delivery Readiness, Evidence Confidence and Trustworthiness), applies six mandatory progression gates, and turns the result into an Executive Canvas with a recommendation, risk statement and remediation plan.
What decisions does I-AVAM™ 2.0 support?
Whether an AI initiative should move to its next stage: from discovery to experiment, pilot, controlled deployment or scale. After deployment, it supports decisions to continue, expand, modify or retire. Each mandatory gate ends in Pass, Conditional Pass, Remediation Hold or Stop.
Why can't a high score approve an initiative on its own?
Because comparison is not permission. The Scale Decision Score compares initiatives that are eligible; mandatory gates decide eligibility. A strong value estimate cannot offset unlawful use, unmitigable harm, unresolved data rights or missing accountable ownership.
Does passing every gate mean the initiative is funded?
No. Passing the gates establishes eligibility for the next decision. Portfolio selection then considers resource availability, cost, dependencies, sequencing, risk concentration and strategic balance. A ready initiative may still be sequenced later.
Is evidence judged the same way at every stage?
No. Evidence expectations are stage-sensitive. Discovery needs problem definition and strategic relevance; scale needs stable outcomes, approved controls and a validated benefit case. The Evidence Confidence Ratio measures completeness for the current stage.
Does I-AVAM™ 2.0 replace NIST AI RMF, ISO/IEC 42001 or the EU AI Act?
No. It is built to use the evidence those governance processes produce. Regulatory permissibility and mandatory assurance obligations enter I-AVAM™ as eligibility conditions rather than as points in an average.
Are the weights and thresholds validated?
Not yet. Equal weights of 0.25 and provisional thresholds are transparent starting assumptions. They will be calibrated through expert review, reliability testing, retrospective and prospective studies, and sensitivity analysis.
How does I-AVAM™ relate to PYRALO™?
I-AVAM™ 2.0 helps leaders decide whether an AI initiative is ready for its next level of commitment. PYRALO™ helps them make sure people adopt it responsibly once it is deployed.