Most retail boards are asking the wrong question. They convene the offsite, they debate GPT versus Gemini versus a fine-tuned open weight model, they argue build versus buy. They treat this as a strategy problem.
It is not. It is an ownership problem. And the sooner your board admits that, the sooner AI stops being a line item and starts being margin.
The evidence is now hard to ignore. Most organizations run a contested field of platforms, each claiming to be the primary AI layer; few could confidently detect a model drifting or failing in production; and the single most-cited barrier to control is the absence of any one owner accountable for AI across the stack. That is not a technology diagnosis. That is an org chart diagnosis.
The tradeoff nobody owns
Here is what actually happens inside a large retailer today. Marketing stands up a personalization agent tuned to lift conversion. Merchandising runs a pricing and markdown agent tuned to protect gross margin. Supply chain deploys a replenishment agent tuned to service level and working capital. Three agents. Three targets. Three VPs who can each walk into a QBR and report green.
And total enterprise margin still goes backward.
The promo agent floods demand on a SKU the replenishment agent cannot cover, so you eat expedite freight and stockouts. The pricing agent protects margin on an item marketing just spent acquisition dollars driving traffic toward. Every agent is locally optimal. The system is globally dumb. Nobody owns the tradeoff between them, because the tradeoff does not live on anyone’s scorecard.
Model selection did not cause that. Your operating model caused that. You wired autonomous decision-makers into functional silos that were designed for humans who at least talked to each other in a Monday meeting. The agents do not go to the Monday meeting.
This is why I keep telling boards: the operating model decides outcomes far more than the model does. You can buy the best retail-native platform on the market and still lose money if no single seat owns the cross-functional tradeoff. Capability without accountability is just faster mistakes.
Name the seat, then give it teeth
The instinct is to appoint a Chief AI Officer and declare victory. Titles are proliferating fast. IBM’s 2026 CEO study found that 76% of surveyed organizations now have a Chief AI Officer, up from 26% in 2025, while 85% of respondents said all functional leaders must become technology experts in their own domain. A title is not accountability. Most of these seats are science-fair curators with a budget and no P&L.
Worse, oversight is often claimed but not real. Even where a central team is claimed, the named owner is most often the general technology executive rather than a dedicated AI authority, and the governance function exists more often as an org-chart aspiration than an operating reality.
So name the seat properly. Call it the Agentic Operating Owner. Not a scientist, not a platform buyer. An operator who sits above merchandising, marketing and supply chain and owns exactly one thing the functions cannot: the tradeoff.
Give that seat three non-negotiable authorities. First, the P&L for cross-functional agent decisions, so the tradeoff shows up on one scorecard instead of falling between three. Second, the authority to set and revoke autonomy limits per agent, per domain, in real time. Third, a kill switch that actually works. If the owner cannot turn an agent off on a Tuesday afternoon without a change-review committee, they do not own it. They observe it.
This matters because scale is arriving before control does. The core problem is that authority was granted to a system before adequate oversight was established, and scale grew before accountability was in place. In a business where a pricing agent can reprice thousands of SKUs before lunch, an ownership vacuum is not a governance gap. It is a balance-sheet event.
The cost of doing nothing is not zero
Boards love to frame the ownership decision as something they can defer until the technology settles. Wait for the standards. Wait for the regulation. Wait for the pilot to prove out.
That framing hides the real number. The Cost of Doing Nothing here is not a stalled roadmap. It is agents optimizing against each other at scale, every day, while your functional leaders each report success. It is markdown dollars leaking into freight, acquisition dollars fighting your own pricing logic, and no single person who can even see it, let alone stop it. Doing nothing is not neutral. It is actively expensive, and it compounds.
The retailers who win the next three years will not be the ones with the best model. Everyone gets access to roughly the same frontier. They will be the ones who decided, early and explicitly, who owns the tradeoff and gave that person a kill switch.
So stop tuning your AI strategy. Name the seat. Fund the authority. If you cannot answer who owns the tradeoff in one sentence, that silence is your strategy, and it is already costing you margin.