cloudwaypoint ← Back to the overview

Add-on option — Phase Zero companion

AI readiness and token economics

Cloud readiness is the foundation for governed AI. The model is rarely the first constraint.

How it runs

Two phases. The first needs no room; the second is brokered.

Initial specification

Specified against zero trust and RMF preparation.

  • Runs against evidence already captured — no second discovery pass.
  • Produces the request package that names what is genuinely missing.
  • Needs no facilitated session, which is what makes it the affordable tier.

Capture session

One bounded conversation — brokered, never impersonated.

  • Runs from mission friction to an accountable next gate without impersonating the data, AI, security, or authorizing authority.
  • No score. No automatic finding. Drafts stay in a local workpaper until deliberately exported.
  • Provider-neutral: the client's own approved AI, an offline local model, or a manual read.

Facilitator authority boundary. Phase Zero due diligence only — no AI authorization, technical design, model selection, or conformity determination. Never paste controlled records, secrets, PII, PHI or raw mission data. Exports are consultant workpapers.

The translation layer

Ten domains become three non-scored lenses.

Evidence is assessed once, then read three ways. Each lens prints its own limit.

Ten assessment domains translated into three AI lenses One band listing the ten core assessment domains feeds three lens boxes: Enable for CSP and mission data, Govern for AI RMF and guardrails, Sustain for FinOps and token economics. Each lens carries a limit statement. One evidence spine — assessed once Governance · platform · operations · service management · DevSecOps · RMF evidence · control inheritance · applications and data · migration · FinOps 01 · Enable CSP + mission data Can a bounded use case reach usable, releasable, attributable data? 02 · Govern AI RMF + guardrails Are use, ownership, data rights and authority clear enough to proceed? 03 · Sustain FinOps + token economics Can consumption be allocated to a mission, workload or outcome? Pattern fit — not provider ranking Awareness — not certification Visibility — not a savings guarantee

The example workflow

Four stages per lens, ending in a name and a gate.

Three lenses, four stages each A grid with three lens rows and four stage columns: observed evidence, AI implication, recommended action, and owner with gate. 01 · Observed evidence 02 · AI implication 03 · Recommended action 04 · Owner / gate Enable csp-mission-data Data location, tenancy and provider boundary. What that means for a model reading mission data. A step that can be scheduled and funded. Named owners and the gate it must pass. Govern ai-rmf-governance Authorization posture and control inheritance. Whether the component lands inside the boundary. The RMF preparation the workload needs first. The authorizing official, named not implied. Sustain finops-token-econ Cost visibility and chargeback maturity. Whether per-token spend can be attributed at all. What must change to govern an AI spend line. The budget owner and the threshold that returns to them. The fourth stage is a name and a gate, not a recommendation. A workshop that ends without those two has produced a reading list.

Mission-data due diligence

Six questions about candidate providers — not a data design.

The six mission-data due-diligence questions Six numbered chips: provider and owner, meaning and lineage, boundary and handling, interface and operations, strategic data intent, and AI-specific unknowns. 01 Provider and owner Who controls access, release and quality? 02 Meaning and lineage Semantics, provenance, freshness, limits. 03 Boundary and handling Where may it be stored, moved, combined? 04 Interface and operations Repeatable access, or a one-time extract? 05 Strategic data intent Proposed, piloted, or actually operating? 06 AI-specific unknowns Evaluation sets, contamination, feedback.
Identifies candidate providers and questions. It does not designate an authoritative source or replace data owners and SMEs.

AI RMF-aware governance

Four questions, borrowed from a framework rather than invented.

NIST AI RMF as four questions A two by two grid: Govern asks who is accountable, Map asks what the context is, Measure asks what evidence is acceptable, Manage asks how risk will be treated. Govern Who is accountable? Roles, risk tolerance, oversight, escalation. Map What is the context? Purpose, consequences, misuse, limits. Measure What evidence is acceptable? Criteria, testing, uncertainty, challenge. Manage How will risk be treated? Mitigation, monitoring, rollback, retirement.
An awareness and due-diligence reference — not a conformity determination.

FinOps for AI

Tokens are a consumption signal — not the unit of mission value.

The AI cost stack, five layers Five stacked bands. From the top: accepted mission or business outcome; agent and workflow behavior; model tokens; retrieval and platform services; cloud infrastructure and labor. Accepted mission or business outcome The only layer that is value Agent and workflow behavior Model input, output, cached and reasoning tokens The layer everyone prices Retrieval, vector, data transfer, tool, API, platform Cloud infrastructure, security, observability, labor
Allocate by mission, product, workload, environment, owner. Observe volume, latency, quality, reliability, risk, unit cost. Decide to continue, optimize, constrain, redesign or retire. This establishes visibility, not a savings guarantee.

What the session ships

Six artifacts. Two of them assume you send nothing anywhere.

The six artifacts the capture session generates Client request package as DOCX, request email as TXT, participant starter pack as Markdown, facilitator runbook as Markdown, transcript review prompt as TXT, and local AI prompt pack as TXT. DOCX Client request package Evidence request and preparation guide. TXT Request email Purpose, minimum evidence, boundaries. MD Participant starter pack Roles, session contract, prompts. MD Facilitator runbook Opening script, seven moves, closing. TXT Transcript review prompt Provider-neutral: approved client AI, or manual. TXT Local AI prompt pack Offline: Ollama or another local model.
The transcript review prompt and local AI prompt pack are why this works inside a restricted estate — neither requires handing anything to an outside service.

Jump-start path

From awareness to one bounded opportunity.

The four-step jump-start path Phase Zero readiness perspective, then a focused workshop for opportunity due diligence, then a specialist motion for design and evidence, then authorized action to proof, implement or hold. Phase Zero Readiness perspective What is known, and what is not. Focused workshop Opportunity due diligence One outcome, bounded. Specialist motion Design and evidence Architecture, controls, FinOps. Authorized action Proof, build or hold Under named authority only. The readiness perspective and focused workshop are the add-on. Later work begins only through an explicit client decision.

This does not produce an AI maturity score, and it never will. The governed workflow is designed and tested to keep the trace non-scored. What it produces is a traced position — what was observed, what it implies, what to do, and who decides — which is auditable a year from now in a way a score is not.

Primary public references

Verified 12 August 2026.

Representative of the companion's structure, not an engagement output. Priced separately from the base engagement. Cloud Waypoint does not grant authorization, certify compliance, replace accountable officials, or select a provider or model. Evidence informs; named individuals with authority decide.