Agent Decision Preference Stack

Definition: The three-layer hierarchical model describing how autonomous AI purchasing agents process and prioritise information when making purchasing decisions on behalf of human consumers.

The Stack comprises three layers, processed in sequence:

Layer 1: Hard Constraints — Non-negotiable requirements set by the consumer (price ceiling, delivery deadline, specifications, certifications). These operate as Boolean gates: a product either passes or is eliminated. Marketing cannot influence this layer.

Layer 2: Learned Heuristics — The agent's accumulated knowledge about brands, product quality, and category reputation, drawn from training data, web content, reviews, and structured knowledge bases. This layer determines which products enter the agent's consideration set. Generative Engine Optimisation (GEO) operates at this layer.

Layer 3: Real-Time Optimisation — Live evaluation of operational performance: current pricing, verified inventory, delivery reliability, structured data completeness, and computational trust signals. This layer determines which product is selected and purchased. Agent Intent Optimisation (AIO®) operates at this layer.

Empirical simulation (600 purchase decisions) found that balanced agents demonstrated 4.71× operational dominance — placing 4.71 units of decision weight on Layer 3 performance for every 1 unit placed on Layer 2 reputation.

Introduced in: Accornero, P.F. (2026). "AGENTIC COMMERCE: A Theory of Markets When the Shopper Is No Longer Human." SSRN #6111766.

Related concepts: Agent Intent Optimisation (AIO®) | AIO vs GEO | Four Ds Framework™