Assess AI product operating-model maturity across seven categories using evidence, disagreement, and critical-gap logic. Use to identify consequential gaps and next interventions.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Assess AI product operating-model maturity across seven categories using evidence, disagreement, and critical-gap logic. Use to identify consequential gaps and next interventions.
["Establishing an evidence-based baseline before operating-model design","Identifying the weakest consequential organizational conditions","Comparing leadership intent with product-team and governance practice","Preparing an AI product operating-model roadmap"]
evidence_required
["Existing strategy, portfolio, workflow, context, evaluation, governance, and capability artifacts","Examples showing whether documented practices are actually used","Decision records, evaluation results, adoption evidence, and outcome measures where available","Perspectives from leadership, practitioners, governance partners, and affected users where relevant"]
produces
["Seven-category maturity profile","Evidence and confidence ledger","Disagreement and critical-gap report","Prioritized intervention recommendations","Named owners and next evidence-producing actions"]
Use this workflow to assess how an organization currently chooses, builds, evaluates, governs, and learns from AI-powered products and AI-assisted product work across seven connected categories.
The assessment changes the organization from unexamined maturity claims toward an evidence-based profile with consequential gaps, named owners, and next interventions. The score is navigation; the changed decision and organizational motion are the point.
Why Use It
Organizations commonly confuse AI activity with capability and documentation with adoption. A single radar chart can also hide different operating levels, weak evidence, serious disagreement, or a missing non-negotiable control.
This workflow helps leaders and teams:
Distinguish discussed, documented, used, and outcome-improving practices
Compare leadership intent with practitioner and governance experience
Separate organization maturity from initiative readiness
Identify the weakest consequential condition rather than the lowest arithmetic score
Prevent strong low-risk scores from offsetting critical gaps
Route findings to concrete skills, owners, and evidence-producing actions
Establish a baseline that can be revisited without rewarding performative activity
When to Use It
Use the assessment before an operating-model design sprint or roadmap, during strategic planning, after rapid growth in AI initiatives, when governance and product teams disagree about readiness, or when leaders need to determine why pilots are not becoming reliable capability.
Do not use the organization assessment as a launch approval for one initiative. Use an initiative-readiness review for that decision while keeping the two profiles separate.
What It Produces
The assessment report includes:
Scope, operating levels, and participants
Evidence inventory and limitations
Scores and confidence for all 35 questions
Seven category profiles
Material disagreements and missing perspectives
Critical gaps that constrain action
Weakest consequential conditions
Prioritized intervention recommendations
Named decision and action owners
Next evidence-producing actions and unresolved questions
Who Should Participate
Build a cross-functional evidence group rather than relying on a single executive workshop.
Include appropriate representatives from:
Product and technology leadership
Product Managers, Team Leads, and product teams
Product Operations
Design, research, data, and engineering
Legal, privacy, security, safety, risk, compliance, or AI governance
Business or operational owners
Affected users, customers, or workforce representatives when relevant
Record which perspectives are missing. Do not treat attendance as evidence of maturity.
Evidence to Bring
Invite artifacts and examples across all seven categories:
Strategy documents and decisions they changed
Portfolio inventories, charters, stage gates, and stop decisions
Workflow maps, human-AI responsibility agreements, and before-and-after measures
Source maps, context packages, data assessments, and lifecycle controls
Behavior contracts, evaluation sets, scorecards, and production reviews
Accountability records, autonomy boundaries, control tests, and incidents
Competency maps, applied learning evidence, reuse records, and outcome measures
Evidence is an invitation, not a gate. When evidence is missing, record the gap, lower confidence, and recommend how to obtain useful evidence. Use Not assessed when the available context cannot distinguish absent practice from unobserved practice. Use 1 — Absent only when inquiry or explicitly attributed reports support absence in the stated scope. Unsupported claims can never score above 2 — Emerging.
How to Do It
1. Set scope before scoring
Name the organization, business unit, geography, product area, time period, and operating levels being assessed. Record the decision the assessment must enable.
Do not combine materially different scopes into one score. Create separate profiles when enterprise policy, portfolio practice, team behavior, or initiative readiness differ.
2. Establish the participant and evidence map
Identify represented and missing perspectives. Inventory supplied artifacts and classify claims as discussed, documented, used, or impact-demonstrated.
Create an initial evidence ledger with source, owner, date, relevance, limitations, confidence, and decision implication.
3. Select the assessment path
Use all 35 canonical questions for a complete baseline. For a phased assessment, work category by category but preserve one shared context ledger and do not claim full-library coverage until all categories are complete.
4. Evaluate each question
For every canonical question:
State the claim being assessed.
Review evidence and counterevidence.
Identify scope and representativeness.
Record disagreement.
Assign evidence strength.
Decide whether a score is supportable; otherwise record Not assessed and the evidence needed.
Select the highest maturity score fully supported.
Assign confidence and explain the limiting factor.
Flag any critical gap.
Name the next evidence-producing action.
Do not ask participants to self-score an unfamiliar concept without explaining the maturity anchors and examining evidence.
5. Build category profiles
Show all five question scores, evidence confidence, disagreement, and critical gaps. If a category summary is needed, use the median and display the underlying distribution.
Do not use a category average to conceal an important weakness.
6. Apply critical-gap logic
Report missing legal, privacy, security, safety, accountability, oversight, evaluation, or rollback conditions separately. State which actions they constrain, such as launch, scale, sensitive-data use, or increased autonomy.
7. Find the weakest consequential conditions
Prioritize conditions based on their effect on decisions and outcomes, not merely their numeric rank. Look for dependencies: a weak source system may undermine evaluation; unclear accountability may prevent a safe autonomy decision; a missing strategy may create a portfolio of unrelated pilots.
8. Recommend interventions
For each priority condition, recommend a specific AIPOM skill or foundational method, explain why it fits, identify the owner, and state the evidence the intervention should produce.
Limit the first improvement horizon to a coherent set of motions the organization can own. Do not prescribe all 41 skills.
9. Confirm the human decision and handoff
The accountable leader decides which findings and interventions to adopt. Record accepted decisions, deferred decisions, disagreements, owners, dates, unresolved questions, and the trigger for reassessment.
Facilitation Protocol
Guided mode
Ask one consequential question at a time. Begin with scope and the decision to be enabled, then move through the canonical questions while reusing the evidence ledger. Explain the maturity anchors before requesting judgment.
Context-dump mode
Read all supplied materials first. Pre-populate claims, evidence, possible scores, confidence, and gaps. Ask only about ambiguities or missing information that could change the profile or recommendation.
Best-guess mode
Produce a provisional profile using labeled assumptions. Cap unsupported scores at 2 — Emerging, leave genuinely unobservable conditions Not assessed, use low confidence, identify missing perspectives, and propose the smallest evidence-gathering actions needed before consequential decisions.
Multi-session handling
At the end of each session, preserve:
Categories and questions completed
Evidence reviewed
Provisional scores and confidence
Disagreements
Critical gaps
Open questions
Next required input
Resume from this ledger without repeating completed questions.
At meaningful routing decisions, present numbered intervention options, explain fit, recommend a default when evidence supports it, and accept combined or custom selections.
Decision Logic
If the requested scope mixes organization maturity and initiative readiness, separate the profiles.
If only executive opinions are available, treat them as claims and run best-guess mode with low confidence.
If a practice is documented but use is not demonstrated, do not score it above 2 — Emerging.
If current examples show repeatable use in part of the scope, consider 3 — Repeatable while stating the boundary.
If organization-wide use is claimed without representative evidence, lower the score or confidence.
If evidence demonstrates measured improvement and reuse, consider 5 — Compounding only within the supported scope.
If respondents disagree, preserve the distribution and investigate operating level, scope, recency, incentives, or missing perspectives.
If a critical gap exists, state the constrained decision before discussing averages.
If several gaps share a root cause, recommend the root intervention before treating symptoms.
If the organization has little evidence across categories, prioritize a small assessment-to-action loop rather than a comprehensive transformation program.
Do not calculate an overall score when scopes differ materially. If stakeholders require one and scopes align, use the median of category medians, label it as a navigation aid, and display all category profiles and critical gaps beside it.
Completion Criteria
Complete the workflow only when it contains:
Defined scope and operating levels
All intended canonical questions assessed
Evidence and assumptions
Score, confidence, and rationale for each question
Seven category profiles for a complete assessment
Preserved disagreements and missing perspectives
Critical gaps and constrained decisions
Weakest consequential conditions
Prioritized interventions with fit rationale
Named owners and next evidence-producing actions
Unresolved questions and reassessment trigger
Do not finish with only scores, a radar chart, or generic recommendations.
Key Concepts
Evidence Before Score
The score summarizes the strongest claim the evidence supports. Confidence does not replace evidence.
Profile Before Average
The distribution across categories and questions is more useful than a single number because operating models fail through consequential weak links.
Consequence Before Arithmetic
The priority condition is the weakness that most constrains responsible value, not automatically the numerically lowest item.
Disagreement Is Data
Different perspectives may reveal policy-practice gaps, uneven adoption, scope differences, or competing risk tolerances.
Assessment-to-Action Loop
An assessment is useful only when it changes a decision, routes work to an intervention, assigns ownership, and produces new evidence.
Organizational Applications
Establish an enterprise or business-unit baseline
Diagnose why pilots do not become repeatable capability
Prepare an operating-model roadmap or design sprint
Compare documented governance with product-team practice
Identify capability and context dependencies across a portfolio
Reassess progress after a defined improvement horizon
Support leadership discussion without manufacturing false consensus
Common Pitfalls
Using executive self-ratings as evidence
Averaging away a critical gap
Mixing organization maturity with initiative readiness
Treating a policy, template, workshop, or license as adoption
Asking all questions mechanically despite supplied context
Hiding disagreement to create a cleaner chart
Recommending every skill instead of the next coherent motions
Confusing completed assessment output with changed organizational practice
Reassessing activity measures rather than decision or outcome evidence
Combine With
aipom-operating-model-roadmap to sequence findings into 30-, 90-, 180-, and 365-day interventions
The seven category advisors to deepen the most consequential category diagnosis
aipom-autonomy-boundary-designer when unclear AI authority is a priority governance gap
aipom-strategy-thesis-advisor when portfolio incoherence begins with a missing strategic thesis
Foundational problem-framing, system-mapping, stakeholder-mapping, and premortem skills when those methods already solve the underlying problem
National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), published January 26, 2023. Supports lifecycle-based, organizational AI risk management; governance; measurement; and documented accountability. Accessed July 16, 2026. NIST notes that AI RMF 1.0 is being revised.
OECD.AI, Advancing Accountability in AI. Supports role- and context-aware accountability and iterative risk governance across the AI lifecycle. Accessed July 16, 2026.
