Target's AI Edge: It's Not the Models, It's the System

Target's Senior Vice President Siobhán Mc Feeney emphasized that the company's true competitive advantage in artificial intelligence lies in the intricate systems and architecture surrounding AI models, rather than the models themselves. This foundational structure is considered Target's "moat," enabling scalable investment in appropriate models for specific challenges.

Mc Feeney detailed a deliberate approach to developing AI agents, beginning with a critical question: what problem are they solving? This evaluation determines if an agent is necessary and, if so, its type, ensuring that solutions are not duplicated and that agents are registered and certified. Key considerations include defining triggers for agent action, establishing lineage for transparency, and carefully managing autonomy levels.

AI agents at Target must earn their autonomy, progressing through a four-level ladder from observation to full end-to-end operation with a human in the loop. Performance is rigorously measured against intended goals, allowing for continuous evaluation and improvement. This meticulous process ensures agents remain accurate and valuable, fostering a culture where builders can innovate rapidly within clearly defined guardrails. The evolving landscape also requires new skill sets for teams managing both human workers and AI systems.