Cluster
Delegation
The delegation cluster covers principles that help teams move from direct manipulation to well-bounded task assignment. Delegation-first design makes the scope of AI authority legible, steerable, and correctable without requiring users to monitor every step.
Key Facts
- Cluster
- Delegation
- Principles
- 3
- Linked examples
- 6 implementation library examples
- Principle titles
- Design for delegation rather than direct manipulation; Replace implied magic with clear mental models; Optimise for steering, not only initiating
Principles in this cluster
- Principle 1Design for delegation rather than direct manipulationDesign experiences around the assignment of work, the expression of intent, the setting of constraints, and the review of results, rather than requiring users to execute each step manually.In agentic systems, value is created when users can define the desired outcome and rely on the system to carry out appropriate actions within agreed limits. The interface should therefore support delegation as a first-class interaction model.
- Principle 5Replace implied magic with clear mental modelsThe product should help users understand what the system can do, what it is currently doing, what it cannot do, and what conditions govern its behaviour.Trust is strengthened when users can form accurate expectations. Systems that appear intelligent but remain poorly bounded create confusion, misuse, and misplaced reliance.
- Principle 10Optimise for steering, not only initiatingThe system should support users not only in starting tasks, but also in guiding, refining, reprioritising, and correcting work while it is underway.Prompting is an initiation mechanism. It is not, by itself, a sufficient control model for complex or consequential work. Users require the ability to steer ongoing activity without restarting the entire process.
Implementation examples
- Level 1: Augmented LLM — Single API CallOne model call with structured output, system prompt, and context. No loops, no tools.script · advanced
- Level 2: Prompt Chains & Routing — Deterministic DAGsMultiple LLM calls in a fixed sequence. Code controls the flow, not the model.script · advanced
- Level 3: Tool-Calling Agent — Scoped AutonomyThe agent decides which tools to call and in what order, but only within a fixed set of well-defined capabilities.script · advanced
- Level 4: Agent Harness — Full Runtime AccessGive the agent a full runtime via the agent runtime SDK. It can search files, read docs, and reason through problems autonomously.script · advanced
- Level 5: Multi-Agent Orchestration — Delegated AutonomyAn orchestrator delegates to specialized subagents defined via the agent runtime SDK. Each subagent has its own prompt, tools, and model. The orchestrator coordinates.script · advanced
- Intelligence: The "brain" that processes information and makes decisions using LLMs.This component handles context understanding, instruction following, and response generation.script · intermediate