Target’s senior vice president of product engineering and data science, Siobhán Mc Feeney, stated the retailer’s advantage in AI lies not in the models but in the systems surrounding them.
“There’s a lot in it. That to us is the advantage,” Mc Feeney said at VB Transform 2026. “The models are useful, but they’re not enough to stand out.”
The discipline behind Target’s AI agents
Target avoids creating AI agents for every issue. Mc Feeney described the current trend toward enterprise AI agents as contentious, noting that not every task requires one. Instead, agents gain autonomy over time rather than getting it by default, shaping how the company designs and implements them.
Related: Waymo’s Approach to AI Evaluation and Safety
She emphasized solving problems that deliver the most value for customers. “We want to make sure we’re investing in the right places.”
Agents are now integrated into Target’s infrastructure, linking signals across supply chain, replenishment, and demand forecasting. Mc Feeney framed this as retail’s oldest promise—the right product, in the right place, at the right time—delivered at scale.
Target’s process is methodical. Before developing an agent, teams evaluate the problem, whether an agent is necessary, and what type of agent would work best. Options include an orchestrator, a domain-specific model, or another solution.
“You define that upfront, and this may sound a little process-heavy, then you have to register and certify your agent,” Mc Feeney said. The process prevents duplicate efforts and establishes clear records for troubleshooting.
Related: EU softens climate rule to permit more pollution
Creating an agent raises additional questions: What triggers it? Is it automated, manual, or time-based? What data and systems does it access? How is it monitored?
“We’re trying to make sure we have lineage from the very beginning—the birthing of this agent, all the way through—because at 2 a.m. one morning, when something goes sideways, we want to make sure we understand everything that happened.”
Agent autonomy and monitoring
New agents typically start with base autonomy and earn more over time. What the agent has access to—data, systems, tables, or databases—is carefully considered.
A digital-twin simulation this summer predicted men’s shorts inventory across three Target stores in Long Beach. One store required six to seven times more stock than the others. Analysts initially questioned the recommendation, but the system had identified a key factor: the store sat less than two miles from the beach, while the others were 10 to 12 miles inland. The stock sold through.
Related: Samsung Enhances Galaxy Z Flip Cover Screen
Monitoring and observability are critical. “We measure everything: What it was intended to do, its calibration, its trajectory, not just runtime and latency,” Mc Feeney said. This transparency allows agents to be adjusted over time.
“You’re talking about architecture and taxonomy and a data governance layer that absolutely had to be established,” she said. This foundation enables Target to scale and invest in the right models for the right problems.
Balancing models and cost
Models have different strengths for different tasks. Frontier models excel at complex tasks like supply chain optimization but can be cost-prohibitive. “So it’s making sure there’s always a cost benefit,” Mc Feeney said.
