As organizations move beyond copilots toward agentic AI, they are embedding AI directly into business processes to drive faster, more autonomous decision-making across merchandising, pricing, inventory planning, and customer engagement. To understand how retailers perceive this shift, MathCo partnered with HFS Research to survey senior retail leaders across Fortune 1000 organizations.
The findings reveal an important disconnect. Retailers are accelerating investment in embedded AI, yet they still place greater trust in it when human oversight remains part of the decision-making process. The result is a widening gap between where organizations are investing and where they are willing to place their confidence.
If organizations already recognize the value of embedding AI deeper into their operations, what’s preventing them from extending the same confidence to autonomous decision-making? As AI evolves from assisting decisions to executing them, building enterprise trust may prove just as important as advancing the technology itself.
The Investment–Trust Disconnect
According to the survey, when asked about their preferred AI deployment model, 69% of retail leaders indicated a preference for AI embedded within business processes, reflecting a clear shift toward solutions capable of operating with greater autonomy. Rather than functioning solely as assistive tools, these systems are increasingly expected to shape outcomes across the business, from what gets stocked to how it’s priced and delivered.
When asked which AI deployment model they trusted most, however, 54% of respondents expressed greater confidence in AI when human oversight remained part of the decision-making process. By comparison, only 12% reported greater trust in AI operating autonomously within business processes.
This is not simply a matter of differing preferences. It points to a specific hesitation: retailers are comfortable expanding where AI operates, but not yet comfortable expanding how much it operates without them. The investment curve and the confidence curve are moving in different directions, and closing that gap is the real challenge ahead, not just adopting more AI, but earning the right to let it act alone.
Why Trust Lags Behind Investment
Greater responsibility also changes what organizations are being asked to trust. While copilots support decision-making by generating recommendations that employees can review, agentic AI increasingly operates within business processes where its actions influence downstream outcomes. The shift is subtle but significant. Organizations are no longer assessing AI as a productivity tool. They are assessing whether it can participate reliably in operational decisions that shape business performance.
Retail makes this distinction especially important because business decisions rarely exist in isolation. A pricing change alters demand, demand patterns strain or ease inventory, inventory levels determine fulfillment speed, and fulfillment ultimately determines whether a customer stays or leaves. As AI takes on decisions inside these interconnected workflows, a single action can set off consequences several steps removed from where it originated, and often several functions away from the team that owns it. That distance is what makes trust hard to earn: it is not enough for AI to be right in the moment, it must be right in ways that hold up as its decisions move through the business.
This reflects a broader shift in how enterprise AI is evaluated. Model accuracy and technical capability are no longer the whole story. What matters increasingly is whether leaders can trace a decision back to its cause, explain it to a regulator or a customer, and step in before a bad call compounds. As AI assumes greater operational responsibility, the threshold for trust rises with it.
Building Enterprise Trust for Autonomous AI
Higher levels of autonomy demand higher levels of organizational control. As AI takes on a larger role inside business processes, retailers need mechanisms that let them set clear limits on what the system can decide alone, catch errors before they reach the customer, and recalibrate those limits as the business or the risk environment changes. Without that control, scaling autonomous AI introduces uncertainty that few organizations are prepared to accept.
Governance, transparency, accountability, and appropriate human oversight are not simply safeguards for autonomous AI. They are what let a retailer put a number on the risk it’s taking with each autonomous decision, assign clear ownership when something goes wrong, and expand AI’s role into more business-critical territory deliberately rather than by accident. Scaling agentic AI, in other words, is as much an organizational readiness problem as it is a technology one.
The role of people doesn’t disappear as AI becomes more autonomous. It shifts from reviewing every output to designing the governance that makes autonomy sustainable. That shift will determine how confidently retailers can move beyond copilots and operationalize agentic AI across the enterprise.
Retail has already demonstrated its confidence in AI’s ability to create business value. What’s still being worked out is how much of the decision-making it gets to own outright, and what has to happen organizationally before that ownership feels safe rather than risky. As agentic AI moves from supporting work to carrying out parts of it independently, the retailers who move fastest won’t necessarily be the ones with the most advanced models. They’ll be the ones who built the trust to let those models act.
For leaders looking to benchmark their position and explore the broader trends shaping enterprise AI, the Take 5 Report, developed in partnership with HFS, offers deeper insights into where the industry stands today and the opportunities that lie ahead.
Click here to read the report.
