The Argument
Agent legibility is the property that makes an autonomous system's state, capabilities, constraints, and reasoning accessible to human understanding. Without legibility, there is no basis for trust calibration. This essay argues that legibility - not transparency - is the design requirement for agentic AI. Transparency asks "can you see what the system is doing?" Legibility asks "can you understand what the system is doing, what it is authorised to do, and what it will do next?" The distinction is not semantic but architectural, and Signal Clarity provides the framework for making agents readable.
The Evidence
Traditional AI transparency - "here is why the algorithm made this recommendation" - is necessary but insufficient for agentic systems. Agentic systems do not merely recommend; they act. The user needs to understand not just what the agent thinks but what it does, what it is authorised to do, and what it will do next. This requires a legibility stack that operates across four dimensions: state legibility (what is the agent doing now?), capability legibility (what can the agent do?), constraint legibility (what is the agent prevented from doing?), and reasoning legibility (why did the agent make this decision?).
The legibility challenge intensifies in multi-agent environments. When multiple agents interact - as in B2B agentic commerce or complex supply chains - each agent must be legible not only to its human principal but to the other agents it transacts with. This creates a compound legibility requirement: agents must be readable by humans and by machines simultaneously, using different signal formats for different audiences. The design of dual-audience legibility is one of the most complex challenges in agentic experience design.
Merchants face a complementary legibility obligation. They must make their products, services, and commercial terms legible to the agents that evaluate and transact with them. This is the merchant-facing dimension of the legibility stack - what the AXD Institute has previously described as Signal Clarity. A merchant's legibility is not about visual presentation but about structured, machine-readable data that agents can parse without ambiguity. The merchants who achieve high legibility scores will be preferentially selected by autonomous purchasing agents.
The Implication
If this thesis is correct, legibility must become a first-class design requirement for every agentic system. Product teams must move beyond "explainability" features - which typically explain past decisions - to prospective legibility: making the agent's future actions, boundaries, and decision criteria visible before they execute. This requires designing legibility into the agent's architecture from the outset, not bolting it on as an afterthought.
Practically, organisations must invest in three capabilities: state communication protocols that convey agent status in real-time, constraint visualisation that makes delegation boundaries visible, and reasoning summaries that translate complex decision processes into human-comprehensible narratives. The organisations that master agent legibility will build the trust relationships that sustain long-term delegation. Those that treat legibility as optional will find their agents trusted only for trivial tasks - never for the high-consequence delegations where agentic commerce creates the most value.