AIAdvanced#Agent Design#Authority#Risk Management
Agenticity
A
Definition
An informal way to describe how much an AI system can observe, plan, call tools, and act. It is not a standardized maturity score, a measure of trustworthiness, or permission to give an agent broad authority.
Source reviewed
- NIST AI Risk Management Framework opens in a new tab
Supports: NIST frames trustworthy AI as context-specific risk management across validity, reliability, safety, security, accountability, transparency, explainability, privacy, and fairness.
- OWASP Top 10 for LLM Applications opens in a new tab
Supports: Untrusted inputs and excessive agency are recognized risks, supporting constrained tool permissions and human review for consequential actions.
Concept Diagram
Key Takeaways
1
Agenticity is informal language for degrees of AI action, not a validated maturity or trust score.
2
Observation, reasoning, tool use, and state-changing execution require separate controls.
3
A model response is not sufficient authority to send, sign, pay, or change an external system.
4
Use least privilege, logs, review, approval, and revocation instead of assuming more autonomy is safer.
Practical Example
An agent reads a protocol's public documentation and produces a draft checklist. It cannot connect a wallet or send a transaction. A separate user reviews the sources and chooses whether to act. This workflow has limited agentic capability and no delegated asset authority.
Finished learning?
Related Terms
AI
Agentic WalletA wallet or smart-account workflow that gives software an explicitly bounded ability to prepare or execute actions. It is an implementation pattern, not an ERC-4337 feature, safety guarantee, or reason to grant a model unrestricted signing authority.
AI
Perception LayerThe part of an AI workflow that collects and labels inputs such as blockchain records, market data, and public communications. It supplies context for review; it does not establish truth, identify people, predict prices, or justify an action.
AI
LLM ReasoningUsing a language model to summarize, classify, or compare supplied information. Its output is probabilistic text, not verified market analysis, financial advice, or authority to execute a trade.
AI
Execution LayerA controlled component that prepares, validates, signs, submits, and monitors a state-changing operation. In an agent workflow, it must treat model output as an untrusted proposal, not as authority to move funds or call a contract.
Related Deep Dives
Master the hub of AI
Explore all our strategic guides about AI to take your operations to the next level.
View all articles