AGENTIC WORK / AIV
AI HAS AGENCY. AGENCY HAS RISK.
Agent Insurable Value estimates the economic exposure generated by authority delegated to autonomous AI. The important question is what the system can do—and what depends on it.
Explore the AIRISK DemoDeveloping exposure framework · Warranty and insurance structures are proposed
AGENT INSURABLE VALUE
Measure the authority. Understand the exposure.
An AI assistant that drafts a purchase recommendation and an agent that can approve the payment may use the same model. Their authority, business dependencies, and potential losses differ.
What goes into the assessment?
- Economic activity entrusted to the system
- Permission to spend, transact, or modify operations
- Dependencies on data, models, vendors, and human oversight
- Limits, monitoring, reversibility, and escalation controls
AIV is an exposure estimate. It is not a quoted premium, a policy limit, an expected claim payment, or the market value of the software.
SEVEN AI-NATIVE RISKS
A practical language for a new kind of agency.
The framework organizes risks into economic, legal, and systemic families.
EconomicRISKS
ARARAutonomous Resource Allocation Risk
An agent provisions cloud resources beyond intended limits.
Examples: Cloud overspend · Runaway GPU usage · Excess API calls
AEDRAutonomous Economic Decision Risk
An agent approves an invoice, changes a price, or negotiates a contract incorrectly.
Examples: Invoice approvals · Procurement · Pricing and trading
AOFRAutonomous Operational Failure Risk
An authorized action disrupts an operation or causes a loss.
Examples: Deleted records · Canceled accounts · Misrouted workflows
LegalRISKS
IPERIntellectual Property Exposure Risk
AI-generated content or code introduces intellectual property exposure.
Examples: Copyright · Trademarks · Code contamination · Trade secrets
PDERPersonal Data Exploitation Risk
The collection, use, or commercialization of personal information creates exposure.
Examples: Unauthorized collection · Consent failures · Regulatory investigations
ILRInformational Labor Risk
Human contributions embedded in AI systems may give rise to economic or legal obligations.
Examples: Training compensation · Collective claims · Unauthorized commercialization
SystemicRISKS
AICRAI Concentration Risk
Dependence on a small number of models or providers creates correlated exposure.
Examples: Model dependency · Cloud concentration · Widespread provider failures
Risk categories help organize assessment evidence. They do not represent seven available insurance products or a determination of legal liability.
INSURANCE READINESS PATHWAY
From assurance to a market conversation.
AI Assurance
Assess governance, controls, documentation, and oversight.
AIV
Estimate economic exposure and make assumptions explicit.
Warranty
Explore proposed contractual performance and risk thresholds.
Insurance
Explore potential residual risk transfer with willing insurers.
READ THE PERSPECTIVE
If AI Compute Is an Investable Asset, AI Agency Must Become Insurable. ↗
James Felton Keith · Insurance Journal · August 21, 2026
THE NEXT STEP
What authority have you delegated?
Start a conversation about your organization, your exposure, and the evidence that matters.