Who
CTO, VP of AI/Data, CISO, Head of Governance, Head of AI/ML Ops, Business Unit Lead, Executive Sponsor
Time Required
60–90 minutes for the full assessment, plus 30 minutes for gap analysis and sprint planning
What You Need
Current AI/agent inventory, infrastructure documentation, governance policies, security assessments, org charts
Output
A7 Readiness Score (7–35), mapped Autonomy Level (L0–L4), gap analysis, and phased readiness sprint
Complete the Organization Profile to establish baseline context.
Score each of the 7 dimensions (A1–A7) using the rubric. For each dimension, identify the level that most closely matches your organization's current state. Write a 1–5 score in the score box.
Calculate your A7 Readiness Score by summing all 7 dimension scores.
Map your score to an Autonomy Level (L0–L4).
Run the Dimensional Floor Check to verify no single dimension blocks your mapped level.
Complete the Agent Washing Detector to validate your current agent deployments.
Perform the Gap Analysis to identify your weakest dimensions and highest-impact improvements.
Build your Agentic Readiness Sprint — a phased plan to move from your current level to the next.
The A7 score is not a grade. It is a diagnostic. A score of 18 does not mean "bad at AI." It means "Copilot Ready" — deploy accordingly. The most expensive mistake in enterprise AI is deploying at an autonomy level your organization cannot support.
Brand commitments behind this assessment
"Not yet" is a first-class outcome. The point of A7 is not to score L4 in three sprints. It is to know exactly where you stand and where the floor blocks you. A diagnosis that says "not yet, here is why" is a successful A7 outcome — often more useful than a marginal climb to the next score band.
L1/L2-only delivery posture. Across every industry segment we measure, no organization scores consistently above L2. L3 is rare; L4 is aspirational. AskAjay advisory engagements are explicitly scoped to L1–L2 deployments only. The L3 and L4 score bands defined later in this worksheet exist for diagnostic completeness — to identify the gap, not to authorize the deployment.
1. Organization Profile
Company Information
Organization Name
Industry
Employees
Annual Revenue
Assessment Date
Assessment Lead
Assessment Team
Current AI Maturity
Total AI/ML systems in production
Systems classified as "agents"
AI governance structure in place?
Dedicated AI/ML team size
AI budget (annual)
Largest AI initiative currently
Current Agent Deployments
List all systems currently classified as "agents" or "agentic AI" in your organization.
Agent / System Name
Use Case
Claimed Autonomy
Vendor / Internal
Status
2. The A7 Assessment
For each dimension, read the five-level rubric and identify the description that most closely matches your organization's current state — not your aspirational state or roadmap. Write the corresponding score (1–5) in the score box. Provide brief justification and evidence.
A1Data Architecture
Score:
/5
Can your data infrastructure serve agents in real-time with the business context they need to make decisions?
1
Siloed
Data in departmental silos. Batch processing only. No API access to operational data. AI training data manually extracted. No data catalog or lineage.
2
Consolidated
Basic data warehouse or lake exists. Some API endpoints. Inconsistent data quality. Limited metadata. Integration requires manual engineering per use case.
3
Integrated
Integrated data lakehouse with broad API access. Data catalog operational. Core data quality metrics tracked. Near-real-time data for priority systems. Emerging semantic layer.
4
Adaptive
Real-time streaming infrastructure. Comprehensive semantic layer with business context. Data mesh principles: domain teams own and publish data products. Agents discover data via self-service interfaces.
5
Agent-Optimized
Agent-native data architecture: real-time streaming, comprehensive semantic layer, agent-specific interfaces, automated quality enforcement, context-rich metadata, feedback loops from agent actions.
Evidence / Justification:
A2Technical Infrastructure
Score:
/5
Does your compute, networking, and orchestration infrastructure support multi-agent coordination, reversible actions, and graceful failure?
1
Basic
Single cloud environment. Manual deployment. No container orchestration. No agent-specific infrastructure. Application-level monitoring only.
2
Standardized
Containerized workloads (Docker/K8s). Basic CI/CD. Cloud-native services. Agents run as standalone processes. No undo capability. No multi-agent orchestration.
3
Orchestrated
Container orchestration with auto-scaling. Agent orchestration framework (LangGraph, CrewAI, etc.). Basic state management. CI/CD includes model deployment. Agent execution monitoring.
4
Resilient
Multi-region/multi-cloud. Sophisticated agent orchestration with tool-use management. Reversible operations for critical actions. Automated failover. End-to-end execution traces. Performance SLAs.
5
Agent-Native
Multi-cloud agent mesh with dynamic routing. Atomic operations with automatic rollback. Auto-scaling for agent workloads. Agent lifecycle management. Self-healing infrastructure.
Evidence / Justification:
A3Governance Framework
Score:
/5
Does your AI governance cover autonomous decision-making, not just model deployment and data privacy?
1
Absent
No AI governance structure. No policies for AI decision-making. No inventory of AI systems. No designated ownership. Agents deployed by whoever has access.
2
Informal
Basic AI policies exist but not enforced. Governance is one person's part-time job. No distinction between traditional AI and agent governance. No decision authority framework.
Cross-functional governance council with authority. Governance in agent development lifecycle. Explicit decision authority per use case. External audit capability. Agent behavior review operational.
5
Embedded
Governance-as-code: agent policies enforced programmatically. Continuous monitoring of decisions against policy. Real-time dashboards. Board-level reporting. Governance enables faster deployment.
Evidence / Justification:
A4Human Oversight Protocols
Score:
/5
Are supervision models, approval workflows, escalation paths, and kill switches in place for agent decisions?
1
None
No oversight mechanisms. No approval workflows for automated decisions. No escalation paths. No kill switches. No monitoring of agent actions.
2
Minimal
Basic action logging exists but not reviewed regularly. Manual intervention possible but requires engineering. Escalation is informal. No self-service kill switch.
3
Defined
Approval gates for critical actions (financial, customer-facing, data modifications). Action logging with regular review. Defined escalation paths. Pause capability exists.
4
Graduated
Graduated oversight: autonomy per agent matched to risk tier. Real-time monitoring dashboards. One-click kill switches with graceful degradation. Automated anomaly detection. Regular effectiveness assessment.
5
Adaptive
Oversight intensity adjusts dynamically based on confidence, novelty, and risk signals. Kill switches with instant graceful degradation. Oversight telemetry feeds governance. Human-agent collaboration protocols. Regular simulation exercises.
Evidence / Justification:
A5Organizational Readiness
Score:
/5
Does the organization have the culture, skills, change management, and executive alignment to deploy and manage autonomous agents?
1
Resistant
No AI culture. Active/passive resistance. No AI skills development. No executive sponsorship. AI perceived as threat.
2
Aware
Executive awareness exists but no strategy or investment. Limited AI skills. Ad hoc change management. AI mentioned but not resourced.
3
Engaged
AI-aware leadership with budget. Skills development programs operational. Dedicated AI team. Change management playbook for AI. Executive sponsor. Cross-functional collaboration.
4
Committed
C-suite aligned on agentic strategy. Dedicated agent operations team. Organization-wide AI literacy. Incentives support adoption. Culture embraces human-agent collaboration. Continuous learning.
5
Agent-Native
Organization designs processes assuming agents are participants. Every function has agent integration expertise. Board-level agent strategy. Contributes to industry standards. Recruits for human-agent collaboration.
Evidence / Justification:
A6Security & Safety
Score:
/5
Does the organization have agent-specific security capabilities beyond traditional application security?
1
Standard AppSec
Traditional AppSec only (SAST, DAST, pen testing). No agent-specific measures. No awareness of prompt injection, tool-use abuse, credential exposure. No agent risk assessment.
2
Aware
Security team aware of agent risks. Basic input validation. Agent credentials managed (not hardcoded). No adversarial testing. Agent security on roadmap.
3
Defensive
Prompt injection defenses. Agent tool-use boundaries defined and enforced. Credential isolation per agent. Output validation for customer-facing actions. Regular security review of configs.
4
Proactive
Agent threat model maintained. Regular adversarial testing / red-teaming. Sandboxed execution for untested actions. Behavioral monitoring. Security in development lifecycle. Cross-agent security assessment.
5
Comprehensive
Full program: scope control, adversarial testing, sandboxing, credential isolation, output validation, behavioral monitoring, supply-chain security for tools/plugins, security automation in agent lifecycle.
Evidence / Justification:
A7Autonomy Calibration
Score:
/5
Can the organization accurately assess its own readiness, distinguish genuine agents from agent-washed products, and deploy the right autonomy level per use case?
1
Uncalibrated
Cannot distinguish assistants, copilots, and agents. No autonomy taxonomy. Vendor claims accepted at face value. "Agent" used interchangeably with "chatbot."
2
Aware
Basic understanding that autonomy levels exist. Informal taxonomy, inconsistent across org. Some vendor claims questioned but no evaluation framework.
3
Defined
Clear autonomy taxonomy adopted organization-wide. Vendor evaluation includes technical assessment. Current deployments accurately classified. Deploying L1–L2 with appropriate oversight.
4
Calibrated
Autonomy matched to readiness per use case. Formal readiness assessment before each deployment. Regular reassessment. Agent washing detector operational — reliably distinguishes genuine agents.
5
Optimized
Right autonomy per use case, continuously reassessed. Clear upgrade paths documented. Organization can articulate what improvements unlock the next level. Contributes to industry autonomy standards.
Evidence / Justification:
3. Score Calculator
A7 Readiness Score
A1 + A2 + A3 + A4 + A5 + A6 + A7 = Total Score
Minimum: 7 | Maximum: 35
A1
/5
A2
/5
A3
/5
A4
/5
A5
/5
A6
/5
A7
/5
A7 Readiness Score/ 35
4. Autonomy Level Mapping
Find your total A7 Readiness Score in the ranges below. Circle or highlight your current autonomy level.
Score 7–14
L0 — Not Ready
Traditional AI only. Deploy predictive models, recommendation engines, dashboards. No autonomous agents. Agents are premature at this readiness level.
Score 15–21
L1 — Copilot Ready
AI assists, human decides. Copilots that suggest actions, draft content, surface insights. A human must approve every action. This is where most organizations are today.
Score 22–28
L2 — Supervised Agent Ready
Agents execute, humans supervise. Agents take autonomous action within defined boundaries. Human oversight monitors decisions and can intervene.
Reality check before the L3/L4 bands
No industry segment we measure scores consistently above L2. The L3 and L4 bands below exist so you can see how far the floor is from your current position — not to authorize deployment. If your score maps to L3 or L4, the right next move is reducing scope until the floor rule clears at the level you can actually support today. AskAjay advisory engagements are scoped to L1–L2 deliverables.
Score 29–33
L3 — Autonomous Ready (diagnostic only — not advisory-delivered)
Agents operate independently within programmatic guardrails. Human review on exception basis. Requires mature governance, robust security, reliable oversight. Most claims of L3 deployment fail the Agent Washing Detector below.
Score 34–35
L4 — Full Autonomy Ready (diagnostic only — aspirational)
Self-directed agents with minimal human oversight. Rare, aspirational. Appropriate only for specific, well-bounded use cases in the most mature organizations.
Your Mapped Autonomy Level
A7 Readiness Score
Mapped Autonomy Level
Current highest deployed level
Premature Autonomy? (deployed > mapped)
5. Dimensional Floor Check
Critical Safety Rule
A single low-scoring dimension can block an autonomy level regardless of total score. L2 requires no dimension below 2. L3 requires no dimension below 3. L4 requires no dimension below 4.
Record each dimension's score below. Check if any score falls below the minimum for your mapped autonomy level.
A1
A2
A3
A4
A5
A6
A7
Floor Rule Matrix
Autonomy Level
Minimum Per-Dimension Score
Implication
L1 (Copilot Ready)
No minimum — any score qualifies
All organizations can deploy at L1
L2 (Supervised Agent)
No dimension below 2
A single Level-1 dimension blocks L2
L3 (Autonomous)
No dimension below 3
Requires Defined/Structured maturity in all 7
L4 (Full Autonomy)
No dimension below 4
Requires Advanced maturity across all 7
Floor Check Result
Lowest dimension score
Which dimension(s)
Floor violation?
Adjusted Autonomy Level (after floor check)
6. Ready for What? — Agent Types by Autonomy Level
Based on your adjusted autonomy level, these are the types of AI systems your organization can safely deploy.
L0 — Not Ready (Traditional AI Only)
Predictive analytics and forecasting models
Recommendation engines (products, content)
Dashboards and BI tools with AI-powered insights
Classification and sentiment analysis
Anomaly detection (alerts, not actions)
L1 — Copilot Ready (AI Assists, Human Decides)
Code copilots (suggest code, human approves)
Email / document drafting assistants
Meeting summarizers and action item extractors
Customer support draft responders (human sends)
Data analysis copilots (suggest queries, human runs)
Multi-agent systems that coordinate without human orchestration
Note: L4 is aspirational for most organizations. Even the most mature firms deploy L4 only for specific, well-bounded use cases.
7. Agent Washing Detector
Why this matters: Gartner estimates only about 130 of thousands of agentic AI vendors are genuine. A product that scores "No" on 3 or more of these questions is likely agent-washed — a rebranded chatbot, copilot, or RPA bot marketed as an autonomous agent.
For each system you have classified as an "agent," answer these five questions.
System Being Evaluated:
1
Planning: Does the system decompose goals into sub-tasks, or does it follow a fixed script?
Yes — decomposes goals
No — fixed script (not an agent)
2
Tool Use: Does the system dynamically select and invoke external tools, or does it call a predetermined API?
Yes — dynamic tool selection
No — predetermined API (not agentic)
3
Memory: Does the system maintain state across interactions and learn from previous actions, or does each interaction start fresh?
Yes — persistent memory
No — stateless (chatbot)
4
Autonomy: Does the system take actions without human approval for each step, or does every action require a click?
Yes — autonomous actions
No — every-click approval (copilot, not agent)
5
Recovery: When the system encounters an unexpected situation, does it adapt its plan, or does it fail and escalate?
Yes — adapts plan
No — always escalates (not autonomous)
"Yes" count
"No" count
Verdict
3+ "No" = likely agent-washed
System Being Evaluated:
1
Planning: Decomposes goals into sub-tasks?
Yes
No
2
Tool Use: Dynamically selects tools?
Yes
No
3
Memory: Persistent state across interactions?
Yes
No
4
Autonomy: Takes actions without per-step approval?
Yes
No
5
Recovery: Adapts plan on unexpected input?
Yes
No
"Yes" count
"No" count
Verdict
3+ "No" = likely agent-washed
8. Gap Analysis
Dimensional Strength & Weakness
Strongest Dimension
Which dimension scored highest? What makes this a strength?
Weakest Dimension
Which dimension scored lowest? What is driving the gap?
Second Weakest Dimension
What is the second-biggest gap?
Biggest Surprise
Which score was most unexpected? Higher or lower than assumed?
Critical Floor Violations
List any dimensions that create floor violations — scores that block your target autonomy level.
Dimension
Current Score
Required Minimum
Gap
Estimated Effort to Close
Highest-Impact Improvement Actions
For each of your bottom 3 dimensions, identify the single action that would have the greatest impact on the score.
Dimension
Current Score
Highest-Impact Action
Owner
Timeline
9. Agentic Readiness Sprint
A phased plan to move from your current autonomy level to the next. Each phase focuses on closing the gaps that block the next level. Do not attempt to skip levels — each level builds on the foundation of the one below it.
Current State
Current Autonomy Level
Target Autonomy Level
Floor violations to clear first
Phase 1Assess + Floor Audit (Weeks 1–2)
Score every dimension. Identify any floor violations that block your current autonomy level. Floor violations mean you are operating above your safe autonomy level — close them first, before any other work.
Identify all floor-violating dimensions
For each: define minimum viable improvement to clear floor
Assign owner and budget for each improvement
Set 30-60-90 day milestones
Re-score affected dimensions at 90 days
Phase 1 Action Items:
Phase 2Remediate the Floor (Weeks 3–12)
With the floor identified, raise the weakest dimensions to enable the next autonomy level. Prioritize by impact: which improvements unlock the most value? Most organizations close their floor violations and stop here — that is the right outcome if your target was L1 or L2.
Identify 2–3 dimensions with biggest gap to next level
Define improvement initiatives with measurable outcomes
Implement governance improvements (A3) if targeting L2+
Build oversight infrastructure (A4) for agent deployments
Conduct agent security assessment (A6)
Run Agent Washing Detector on all "agent" deployments
Phase 2 Action Items:
Phase 3Level Up (Months 4–12)
Build toward your target autonomy level with specific use-case readiness. Deploy agents at the new level for selected use cases and validate the readiness assessment in practice. AskAjay advisory engagements stop at L2 — the L3+ pilot work in this phase is in-house only.
Full A7 reassessment at 6 months
Select 1–2 use cases for next-level pilot deployment
Define success criteria and monitoring plan for pilots
Build organizational learning from pilot deployments
Update autonomy calibration (A7) based on experience
Document upgrade path to next level beyond target
Phase 3 Action Items:
Reassessment Cadence
Quarterly: review actively improving dimensions. Semi-annually: full A7 reassessment. Event-triggered: reassess after major incidents, technology changes, or new agent deployments.
10. Industry Benchmarks
Directional benchmarks based on available evidence as of early 2026. Use for orientation, not prescription. Your organization's score will be unique to its specific capabilities.
Industry Segment
Typical A7 Score
Typical Level
Key Strengths
Key Gaps
Large Tech Companies
22–28
L2
Strong A1, A2 (data & infrastructure)
A3 (governance) and A4 (oversight) often lag technical dimensions
Financial Services
18–24
L1–L2
Strong A6 (security) from regulatory culture
Weak A1 (data) due to legacy systems; A5 (org readiness) mixed
Healthcare
14–20
L0–L1
Strong A3 (governance) from regulatory culture
Weak A2 (infrastructure) and slow A5 (org readiness) adoption
Manufacturing
12–18
L0–L1
Operational discipline (A4 potential)
Weak across most dimensions; A5 (org readiness) is primary blocker
Startups / Digital Natives
16–24
L1–L2
Strong A1, A2 (modern stack)
Weak A3, A4, A5 — move fast, governance lags
Mid-Market Enterprises
10–16
L0–L1
Varies widely
Resource constraints limit all dimensions; focus on L1 readiness first
Your Position vs. Industry
Your industry segment
Industry typical score range
Your A7 score
Position (below / in-line / above)
Key insight from comparison
11. Cross-Framework Integration
The A7 Framework connects to four other frameworks in the AskAjay ecosystem. Each integration point reinforces the others.
Minimum Viable Governance (MVG)Integration: A3 Level 3 = MVG
An organization that has implemented MVG — complete AI inventory, designated owners, risk tiers, deployment gates, monitoring — has achieved Level 3 on A3. If your A3 score is below 3, start with MVG. It provides a 90-day path to structured governance.
askajay.ai/thinking/ai-governance-framework-minimum-viable-governance
The Trust PremiumIntegration: Higher A7 scores → higher Trust Premium
Readiness enables trust. Higher A7 scores (strong data, governance, oversight) naturally produce higher Trust Premium scores — particularly P1 (Risk Avoidance) and P2 (Performance Acceleration). A7 measures capability. Trust Premium measures the value that capability creates.
askajay.ai/thinking/trust-premium-framework-why-trusted-ai-worth-more
The Liability LedgerIntegration: Premature Autonomy → Liability Ledger entries
Every instance of Premature Autonomy generates ethical debt in the Liability Ledger — governance gaps, safety risks, accountability voids. A7 is the preventive mechanism. The Liability Ledger captures the cost when prevention fails.
askajay.ai/thinking/liability-ledger-how-ai-liability-compounds
PRIME FrameworkIntegration: PRIME = build well, A7 = deploy ready
PRIME governs the agent development process. A7 governs deployment readiness. A PRIME-compliant agent deployed into a low-A7 organization still fails — not because the agent is flawed, but because the organizational infrastructure cannot support it.
askajay.ai/thinking (coming soon)
Your Cross-Framework Status
MVG implemented?
Trust Premium Score (if assessed)
Liability Ledger Score (if assessed)
5-Pillar Readiness Score (if assessed)
Key integration insight
12. Notes
Assessment Notes & Observations
Key Decisions & Next Steps
Reassessment Schedule
Next quarterly check-in
Next full reassessment
Reassessment lead
Distribution list
Glossary
A7
The seven-dimension readiness framework: A1 Data Architecture, A2 Technical Infrastructure, A3 Governance, A4 Human Oversight, A5 Organizational Readiness, A6 Security, A7 Autonomy Calibration.
L0–L4
Autonomy levels. L0 Manual (no AI action). L1 Copilot (human approves every action). L2 Supervised Agent (human supervises bounded action). L3 Autonomous (independent within guardrails — rare; not advisory-delivered). L4 Full Autonomy (self-directed; aspirational).
Premature Autonomy
Deploying agents at an autonomy level the organization's infrastructure, governance, and oversight cannot support. Primary driver of the 40%+ agentic-project cancellation rate.
Dimensional Floor Rule
A single low-scoring dimension can block an autonomy level regardless of total score. L2 requires no dimension below 2; L3 requires no dimension below 3; L4 requires no dimension below 4. The weakest link determines the chain's strength.
Agent Washing
Marketing or describing a system as an "agent" when it does not meet the operational definition. Most claimed L3 deployments fail the Agent Washing Detector and are operating at L1 or L2.
"Not yet" as outcome
A first-class A7 verdict. A diagnosis that says "you are not ready yet, here is why" is a successful outcome — often more useful than a marginal climb to the next score band.
L1/L2-only delivery
AskAjay advisory engagements are scoped exclusively to L1 and L2 deployments. L3+ work is in-house only. The L3/L4 score bands in this worksheet exist for diagnosis, not authorization.
Evidence Base
Every claim in this worksheet traces back to the canonical A7 articles, which carry the full citation set. The articles are the source of truth; this worksheet operationalizes them.
Gartner predictions on agentic AI cancellation rates (2025), IDC/WEF governance maturity studies, MIT data architecture research, Microsoft agent-security guidance, McKinsey State of AI 2025 — all referenced inline in the canonical articles.