How to read this evidence
A decision dossier—not a dashboard or sales brochure
The representative scenario shows ten connected outputs. Sources, assumptions, confidence, owners, unresolved evidence and approval boundaries stay visible so an executive can distinguish a signal from an accepted decision.
Decision evidence · baseline
Economic Baseline and Confidence Statement
Representative estate boundary
- Business context
- Global digital-services portfolio with customer-facing SaaS and AI-assisted workflows
- Technology scope
- AWS, Azure, Google Cloud, Kubernetes and two AI-provider usage exports
- Baseline
- USD 2.47m monthly · USD 29.64m annualised
- Allocation
- 72% allocation coverage · 18% shared · 10% unallocated
- Decision window
- Next two planning quarters
- Currency
- USD, synthetic planning values only
Confidence: moderate
Provider totals are 96% reconciled in the synthetic source register. Product allocation is 72% covered; the successful-outcome denominator is 64% evidenced and still requires owner acceptance.
- Allocated product consumption: USD 1.78m monthly
- Shared platform pool: USD 0.44m monthly
- Unallocated evidence backlog: USD 0.25m monthly
Source register
Customer-level allocation, AI task-success denominator, commitment portability and finance acceptance remain open. No value claim advances until the relevant owner approves the evidence.
Accountability · allocation
Allocation and Ownership Map
- ProviderCloud and AI vendorsInvoice and usage origin
- PlatformAccounts, clusters, modelsTechnical control boundary
- ProductSaaS and AI workflowsAllocation destination
- OwnerTechnology + financeDecision and acceptance rights
Direct consumption
Assigned from provider evidence to a product, environment and accountable technical owner.
Shared platforms
Allocated through an approved driver; unallocated remainder stays visible rather than disappearing into averages.
Owner exception
Any material spend without a decision owner enters the evidence backlog before optimisation.
Client approval required: allocation policy, product-value definition and financial reporting treatment remain client-owned.
Unit economics · cloud and AI
Cloud and AI Unit Economics Tree
| Unit layer | Numerator | Denominator | Representative unit | Guardrail |
|---|---|---|---|---|
| Cloud-native product | USD 1.20m allocated cloud | 24.0m transactions | USD 0.050 per transaction | 99.9% service objective |
| AI workflow | USD 0.60m model, data, platform and review | 7.2m successful tasks | USD 0.083 per successful AI task | ≥89% accepted quality; observed 90% |
| Customer economics | USD 1.78m attributed technology | 180k active accounts | USD 9.89 per active account | Finance-approved allocation policy |
The pack does not present a universal unit cost. The unit is only valid after product, finance and technical owners accept its definition and constraints.
Guardrails · decision admissibility
Quality Reliability and Risk Constraint Record
An option that reduces theoretical spend but breaches an accepted quality, reliability, security or commercial constraint is not an admissible optimisation.
Decision portfolio · accountable action
Portfolio Decision Register
Representative decisions show the method. They are not recommendations for any real estate and do not imply realised outcomes.
Planning · scenario comparison
Forecast and Scenario Comparison
| Scenario | Quantified assumption | Monthly planning range | Variance from base | Decision consequence |
|---|---|---|---|---|
| Base | Approved product forecast | USD 2.47m | Reference | Maintain flexibility |
| Growth | Successful-task demand +18% | USD 2.74m–2.82m after qualified decisions | Growth counterfactual: USD 2.91m monthly | Stage commitment after evidence |
| Downside | Demand −15% | USD 2.18m–2.28m | −8% to −12% | Preserve exit options |
| Quality shift | Accepted quality 90% → 94% | USD 2.55m–2.63m | +3% to +6% | Revalidate AI architecture |
Assumptions and exclusions
- No tax, accounting or FX treatment.
- No provider-price forecast.
- No revenue or savings guarantee.
- No production change authority.
Confidence boundary
Scenario confidence depends on workload seasonality, demand ownership, contract terms, architecture feasibility and accepted business units.
Qualified decision range: USD 2.74m–2.82m monthly. Low-confidence synthetic range against the USD 2.91m growth counterfactual, only if the qualified decisions are approved and quality, reliability and security constraints remain satisfied. It is not realised value or a client savings claim.
Commercial exposure · commitment
Commitment Readiness Record
Commitment approval is withheld until demand confidence, portability and authorised owner evidence are complete. AICS does not broker, resell or automatically recommend provider commitments.
Execution · 90-day portfolio
90-Day Action Portfolio
Approve allocation policy
Owner: Technology Finance
Dependency: product and shared-platform mapping
Decision date: day 15
Qualify the AI outcome unit
Owner: AI Product Leader
Dependency: task-success and review telemetry
Decision date: day 30
Test Kubernetes allocation
Owner: Cloud Architect
Dependency: namespace, idle and SLO evidence
Decision date: day 45
Resolve commitment readiness
Owner: CFO Delegate + Procurement
Dependency: demand confidence and provider terms
Decision date: day 60
Decide legacy workload retirement
Owner: Engineering Director
Dependency: business-owner confirmation and rollback
Decision date: day 75
Approve value verification design
Owner: Technology Finance
Dependency: baseline, counterfactual and evidence owner
Decision date: day 90
Dependencies, technical approval, procurement authority and rollback conditions remain visible for every action.
Proof · realised value
Value Realisation Entry
- Observed
- Qualified
- Approved
- Implemented
- Measured
- Finance-accepted
- Sustained
Synthetic planning record—not realised value
- Action: retire the representative legacy analytics workload
- Baseline: USD 86k monthly · synthetic 90-day observed run-rate
- Counterfactual: USD 84k–90k monthly · seasonality range if no action occurs
- Expected-value range: USD 62k–74k monthly cashable reduction · before one-time migration cost of USD 18k–26k
- Evidence state: qualified—not implemented
- Finance acceptance: pending
- Persistence window: 90 days after implementation
Evidence record and counterfactual
Source IDs: CUR-SYN-01, workload inventory SYN-WL-17 and business-owner confirmation pending. The counterfactual uses the synthetic workload's trailing range; it is not an external benchmark.
An opportunity estimate is not delivered value. No amount advances to finance-accepted until implementation, measurement and persistence evidence exist.
Executive brief · unresolved approvals
Executive Decision Summary
Approve evidence resolution and bounded technical qualification
Do not approve a broad commitment or claim value yet.
Material ownership and unit-evidence gaps remain
The estate can be governed without pretending uncertainty has disappeared.
CTO/CIO, finance, FinOps, architecture, product and procurement
Each retains the authority defined for its decision boundary.
Decision workshop with approved source and constraint register
Outcome may be action, hold, referral or no further work.
No billing files, credentials or work email are required to inspect this pack. A review conversation validates fit, decision urgency, stakeholders and scope before any data-access agreement.
Commercial next step
Use the method before sharing sensitive data
Inspect the complete evidence architecture first. If a material cloud or AI economic decision exists, the next step is a fit and scope conversation—not a free audit or an immediate data upload.